Though less well-known than groups like Volt Typhoon and Salt Typhoon, Brass Typhoon, or APT 41, is an infamous, longtime espionage actor that foreshadowed recent telecom hacks.

As China continues its digital gambit around the world, researchers are warning that hacking activity from long-tracked groups is evolving and blending together. On top of that, attackers are hiding their campaigns more effectively and blurring the lines between cybercriminals and state-backed hacking.

Last year, revelations rocked the United States federal government that the Chinese hacking group known as “Salt Typhoon” had breached at least nine major US telecoms. And the group’s rampage even continued into this year in the US and other countries around the world. Meanwhile, the Beijing-linked hacking group “Volt Typhoon” has continued to lurk in US critical infrastructure and utilities around the world. Meanwhile, the notoriously versatile syndicate known as Brass Typhoon—also called APT 41 or Barium—has been operating in the shadows.

The group, which researchers have been tracking since about 2012, has quietly continued its broad targeting around the world over the past year. Brass Typhoon has cast a wide net, leading researchers to view it as a sort of broad coalition that has attacked everything from a US livestock app to source code and chip designs from Taiwan’s semiconductor industry and even power grids. And over the last year, the group has compromised international institutions in the tech and automotive sectors, materials, shipping and logistics, media, and more, using new and refined malware in an array of sustained campaigns.

“They’re absolutely still active and still evolving,” says John Hultquist, who leads threat intelligence at the Google-owned cybersecurity firm Mandiant. “But it’s harder to attribute some of this activity than it was in the past, because it’s all part of a much bigger ecosystem of China’s activity which has been deliberately built to create a tremendous amount of capability.”

Brass Typhoon is known for having carried out a notable string of software supply chain attacks in the late 2010s and for brazen attacks on telecoms around the same time in which the group specifically targeted call record data. The gang is also known for its hybrid activity, carrying out hacks that align with Chinese state-sponsored espionage by the Chinese Ministry of State Security, but also moonlighting on seemingly cybercriminal projects, particularly focused on the video game industry and in-game currency scams.

Research indicates that Brass Typhoon has continued to be active in recent months with financial crimes targeting online gambling platforms as well as espionage targeting manufacturing and energy firms. Its sustained activity has run in parallel to Salt and Volt Typhoon’s recent, attention-grabbing campaigns, and analysis increasingly shows that China’s state-backed hacking operations must be viewed comprehensively, not just in terms of individual actors.

“I think we should not get too down the rabbit hole of is it Salt? Is it Flax? Is it Volt?” former US Cybersecurity and Infrastructure Security Agency director Jen Easterly told WIRED during her last days in that role in January, referring to an array of Beijing-linked hacking groups. “At the end of the day, China, as we've seen in assessments from the Intelligence Community, is the most formidable, persistent cyber threat that we are dealing with.”

Hultquist agrees, emphasizing that while tracking the activity of individual groups is still vital, it is increasingly important for defenders to factor in the advantages that state espionage and offensive hacking operations gain from broad collaboration.

“There was a time when there were very simple indicators that told us who each actor was, and they were operating incredibly loudly, so it was easy to spot the smash-and-grab nature of the activity,” he says. “APT 41 is still doing some loud activity, but so much of its activity now has gotten better and they’ve made an effort to really avoid our controls.”

Ultimately, though, researchers say that the most significant takeaway about Brass Typhoon’s current activity is that it continues apace.

Brass Typhoon: The Chinese Hacking Group Lurking in the Shadows

Despite their hacktivist front, CyberAv3ngers is a rare state-sponsored hacker group bent on putting industrial infrastructure at risk—and has already caused global disruption.

The intermittent cyberwar between Israel and Iran, stretching back to Israel's role in the creation and deployment of the Stuxnet malware that sabotaged Iran's nuclear weapons program, has been perhaps the longest-running conflict in the era of state-sponsored hacking. But since Hamas' October 7 attack and Israel's retaliatory invasion of Gaza, a new player in that conflict threatens not just digital infrastructure in Israel but also critical systems in the US and around the world.

The group known as CyberAv3ngers has, in the last year and a half, proven to be the Iranian government's most active hackers focused on industrial control systems. Its targets include water, wastewater, oil and gas, and many other types of critical infrastructure. Despite being operated by members of Iran's Revolutionary Guard Corps, according to US officials who have offered a $10 million bounty for information leading to their arrest, the group initially took on the mantle of a “hacktivist” campaign.

CyberAv3ngers has been vocal about their operations that targeted Israel and Israeli technology products. But they've also quietly expanded their target list to include a variety of other devices and networks, including a US oil and gas firm and a wide array of industrial control systems across the world.

All of that makes the hackers, despite their grassroots front, a rare example of state-sponsored cybersaboteurs who have crossed the line of targeting and disrupting critical infrastructure. And they haven't shown any signs of stopping.

“They pretend to be hacktivists, but they're really not. This is a state-sponsored group. They have funding and tooling,” says Kyle O'Meara, a threat intelligence researcher at industrial-control-system cybersecurity firm Dragos, which tracks the group under the name Bauxite. “They definitely have the capability, they have the intent, and they have the interest in learning how to shut things off and potentially cause harm.”

Though CyberAv3ngers was active as early as 2020, it first came to prominence in November 2023, after Hamas launched its October 7 attack that killed more than 1,200 people and Israel responded with a ground invasion and bombing campaign that has since killed more than 50,000 Palestinians. A month into that ongoing war, the hackers gained access to more than 100 devices sold by the Israeli firm Unitronics—industrial control systems most commonly used in water utilities and wastewater plants. “Every Equipment ‘Made In Israel’ Is Cyber Av3ngers Legal Target!” read a post from the group's X account.

In that hacking spree, CyberAv3ngers set the names of the devices to read “Gaza” and changed their displays to show an image of the group's logo along with a star of David sinking into ones and zeros. “You have been hacked,” the image read. “Down with Israel.”

While CyberAv3ngers' initial foray may have appeared to be simple vandalism, The hackers actually rewrote the devices' so-called “ladder logic,” the code that governs their functionality. As a result, the hackers’ changes disrupted service on some victim networks, including a water utility and a brewery near Pittsburgh—distinct facilities that were both coincidentally in the same region—as well as multiple water utilities in Israel and Ireland, according to Dragos and another industrial cybersecurity firm, Claroty, that tracked the hacking campaign.

Around the same time, CyberAv3ngers also posted on Telegram that it had hacked into the digital systems of more than 200 Israeli and US gas stations—incidents which Claroty says did occur in some cases, but were largely limited to hacking their surveillance camera systems—and to have caused blackouts at Israeli electric utilities, a claim that cybersecurity firms say was false.

That initial wave of CyberAv3ngers hacking, both real and fabricated, appears to have been part of a tit-for-tat with another highly aggressive hacker group that is widely believed to work on behalf of Israeli military or intelligence agencies. That rival group, known as Predatory Sparrow, repeatedly targeted Iranian critical infrastructure systems while similarly hiding behind a hacktivist front. In 2021, it disabled more than 4,000 Iranian gas stations across the country. Then, in 2022, it set a steel mill on fire in perhaps the most destructive cyberattack in history. Following CyberAv3ngers’ late 2023 hacking campaign, and missile launches against Israel by Iranian-backed Houthi rebels, Predatory Sparrow retaliated again by knocking out thousands of Iran's gas stations in December of that year.

“Khamenei!” Predatory Sparrow wrote on X, referring to the supreme leader of Iran in Farsi. “We will react against your evil provocations in the region.”

Predatory Sparrow's attacks have been tightly focused on Iran. But CyberAv3ngers hasn't limited itself to Israeli targets, or even Israeli-made devices used in other countries. In April and May of last year, Dragos says, the group breached a US oil and gas firm—Dragos declined to name which one—by compromising the company's Sophos and Fortinet security appliances. Dragos found that in the months that followed, the group was scanning the internet for vulnerable industrial control system devices, as well as visiting the websites of those devices’ manufacturers to read about them.

Following its late 2023 attacks, the US Treasury sanctioned six IRGC officials that it says were linked to the group, and the State Department put its $10 million bounty on their heads. But far from being deterred, CyberAv3ngers has instead shown signs of evolving into a more pervasive threat.

Last December, Claroty revealed that CyberAv3ngers had infected a wide variety of industrial control systems and internet-of-things (IOT) devices around the world using a piece of malware it developed. The tool, which Claroty calls IOControl, was a Linux-based backdoor that hid its communications in a protocol known as MQTT used by IOT devices. It had been planted on everything from routers to cameras to industrial control systems. Dragos says it found devices infected by the group worldwide, from the US to Europe to Australia.

According to Claroty and Dragos, the FBI took control of the command-and-control server for IOControl at the same time as Claroty's December report, neutralizing the malware. (The FBI didn't respond to WIRED's request for comment about the operation.) But CyberAv3ngers’ hacking campaign nonetheless shows a dangerous evolution in the group's tactics and motives, according to Noam Moshe, who tracks the group for Claroty.

“We're seeing CyberAv3ngers moving from the world of opportunistic attackers where their whole goal was spreading a message into the realm of a persistent threat,” Moshe says. In the IOControl hacking campaign, he adds, “they wanted to be able to infect all kinds of assets that they identify as critical and just leave their malware there as an option for the future.”

Exactly what the group might have been waiting for—possibly some strategic moment when the Iranian government could gain a geopolitical advantage from causing widespread digital disruption—is far from clear. But the group's actions suggest that it's no longer seeking to merely send a message of protest against Israeli military actions. Instead, Moshe argues, it’s trying to gain the ability to disrupt foreign infrastructure at will.

“This is like a red button on their desk. At a moment's notice they want to be able to attack many different segments, many different industries, many different organizations, however they choose,” he says. “And they're not going away.”

CyberAv3ngers: The Iranian Saboteurs Hacking Water and Gas Systems Worldwide

For the past decade, this group of FSB hackers—including “traitor” Ukrainian intelligence officers—has used a grinding barrage of intrusion campaigns to make life hell for their former countrymen and cybersecurity defenders.

Russian state hackers, perhaps more than those of any other nation, tend to show off. The notorious Sandworm unit within Russia's GRU military intelligence agency, for instance, has triggered unprecedented blackouts and released destructive, self-replicating code. The FSB's ingenious Turla group has hijacked satellite internet connections to steal victims' data from space. But one team of less-flashy cyberspies working on behalf of the Kremlin rarely earns the same notice: Armageddon, or Gamaredon.

The hackers, believed to work in the service of Russia's FSB intelligence agency, aren't known for their sophistication. Yet they have strung together a decade-plus record of nearly constant espionage-focused breaches, grinding away with simple, repetitive intrusion methods, year after year. Thanks to that sheer overwhelming quantity of hacking attempts, they represent by some measures the top espionage threat facing Ukraine in the midst of its war with Russia, according to cybersecurity defenders who track the group.

“They are the most active state-aligned hacker group attacking Ukrainian organizations, by far,” says Robert Lipovsky, a malware researcher at Slovakian cybersecurity firm ESET.

ESET has tracked Gamaredon as it's breached the networks of hundreds of victims in Ukraine, stealing thousands of files on a daily basis, Lipovsky says. “Their operation is highly effective," says Robert Lipovsky, a malware researcher at ESEThe adds. "Volume is their big differentiator, and that's what makes them dangerous.”

If Gamaredon doesn't behave like other Russian hacking groups, that's in part because some of them aren't Russian nationals—or weren't, technically, until 2014.

According to the Ukrainian government, Gamaredon's hackers are based in Crimea, the peninsula of Ukraine that was seized by Russia following Ukraine's Maidan revolution. Some of them previously worked on behalf of Ukraine's own security services before switching sides when Russia's Crimean occupation began.

“They are officers of the ‘Crimean’ FSB and traitors who defected to the enemy,” reads one 2021 statement from the Ukrainian SBU intelligence agency, which alleges the group carried out more than 5,000 attacks on Ukrainian systems including critical infrastructure like “power plants, heat and water supply systems.”

The group's initial access techniques, ESET’s Lipovsky says, consist almost entirely of simple spearphishing attacks—sending victims spoofed messages with malware-laced attachments—as well as malicious code that can infect USB drives and spread from machine to machine. Those relatively basic tactics have hardly evolved since the group first appeared as a threat aimed at Ukraine in late 2013. Yet by tirelessly cranking away at those simple forms of hacking and targeting practically every Ukrainian government and military organization—as well as Ukrainian allies in Eastern Europe—on a daily basis, Gamaredon has proven to be a serious and often underestimated adversary.

“People sometimes don’t realize how big a part ‘persistence’ plays in the phrase APT,” says John Hultquist, chief analyst for Google's Threat Intelligence Group. "They’re just relentless. And that itself can be kind of a superpower.”

In October 2024, the Ukrainian government went as far as to sentence two of Gamaredon's hackers in absentia for not only hacking crimes but treason. A statement from the SBU at the time accused the two men—neither of whom are named—of having “betrayed their oath” by voluntarily joining the FSB.

For Gamaredon's former SBU hackers, turning on their former countrymen may not have resulted in the perks they hoped. Aside from the apparent slog of their nonstop phishing campaigns, intercepted phone communications between members of the group published by the SBU appear to show them complaining about their low pay and lack of recognition. “They should have given you a medal,” one team member says to another in the Russian-language conversation. “Screwed one more time.”

Given how mind-numbingly workaday their hacking campaigns are, it's no wonder they complained about their working conditions, says Google's Hultquist.

"Drudgery is so core to their operations,” he says. "This group grinds out wins."

As disgruntled as Gamaredon's hackers may be, defending against their constant barrage of spying attempts is at least as difficult and boring, say some of the defenders tasked with tracking them. The group writes its malware in relatively unsophisticated scripting languages like VBScript and Powershell rather than the C++ used by savvier hackers. But Gamaredon tweaks its humdrum code constantly, sometimes with automated changes to create endlessly differentiated versions designed to defy antivirus, according to ESET, whose anti-malware products are used widely across Ukraine.

In some cases, the hackers infect the same machine with numerous malware specimens, and hit so many targets that ESET hasn't even been able to identify all of the group's victims, despite closely tracking Gamaredon's campaigns.

“It's exhausting work,” says Anton Cherepanov, an ESET malware researcher. “People overdose and get burnt out.”

Since the start of Russia's full-scale war in Ukraine in 2022, Gamaredon has evolved to broaden its intelligence collection to messaging tools like Signal, WhatsApp, and Telegram, as well as the Delta software used by the Ukrainian military on tablet computers. A 2023 report by CERT-UA, the Computer Emergency Response Team of Ukraine, warned that Gamaredon has on at least one occasion launched a data-destroying attack against a victim facility, though it usually confines itself to mere intelligence gathering on behalf of the Russian military effort.

The same report notes that once Gamaredon infects a machine, it often starts stealing files in as little as 30 minutes. By the end of a week, if the machine remains infected, the hackers will have installed 80 to 120 variants of its malware on the computer. If defenders fail to delete even one, the hackers keep their foothold and can maintain access to that device.

All of that means Gamaredon represents a challenge that's painfully dull for cybersecurity defenders, but with dauntingly high stakes in the context of a war where stolen secrets can mean the difference between life and death.

“They're not interesting,” ESET malware researcher Zoltán Rusnák says. “Just dangerous.”

Gamaredon: The Turncoat Spies Relentlessly Hacking Ukraine

From crypto kingpins to sophisticated scammers, these are the lesser-known hacking groups that should be on your radar.

Read More

_The Last of Us_ Creator Didn’t Want to Sweep ‘Upsetting’ Moments Under the Rug

During _The Last of Us_’ second-season premiere, Ellie gets called a homophobic slur. Craig Mazin tells WIRED it highlights what happens when humanity stops moving forward.

FTC v. Meta Trial: The Future of Instagram and WhatsApp Is at Stake

The blockbuster antitrust case begins Monday. Its outcome could impact how Big Tech companies grow—but the government has a long way to go to prove its case.

The Best Laptops to Work and Play Wherever You Are

These are our favorite Windows laptops, MacBooks, Chromebooks, and Linux portables.

The Best Backpacking Water Filters

We’ve tested filter from Sawyer, Katadyn, LifeStraw, MSR and more to find the best option for every adventure.

The Best Way to Roast a Chicken Is on a Stick

This simple gadget is a fun and inexpensive way to tinker.

Small Language Models Are the New Rage, Researchers Say

Larger models can pull off a wider variety of feats, but the reduced footprint of smaller models makes them attractive tools.

Homeland Security Email Tells a US Citizen to 'Immediately' Self-Deport

An email sent by the Department of Homeland Security instructs people in the US on a temporary legal status to leave the country. But who the email actually applies to—and who actually received it—is far from clear.

The Best Google Pixel 9 Cases and Accessories

Whether you saved some cash with the Pixel 9a or went big with the Pixel 9 Pro XL, we’ve got a selection of cases—MagSafe included—to kit out your new Android phone.

Am I the Asshole if I Don’t Put My Phone on Airplane Mode?

If you don’t, your flight won’t crash, but ignoring airplane mode uses up your battery and annoys the pilot.

The 24 Best Shows on Amazon Prime Right Now

_The Wheel of Time, Reacher_, and _The Bondsman_ are just a few of the shows you should be watching on Amazon Prime Video this week.

China Secretly (and Weirdly) Admits It Hacked US Infrastructure

Plus: The Department of Homeland Security begins surveilling immigrants' social media, President Donald Trump targets former CISA director who refuted his claims of 2020 election fraud, and more.

The Most Dangerous Hackers You’ve Never Heard Of

Millions of scam text messages are sent every month. The Chinese cybercriminals behind many of them are expanding their operations—and quickly innovating.

The scam text messages follow a similar pattern: You need to pay an outstanding toll road fee, or a parcel can’t be properly delivered, they say. “The USPS package arrived at the warehouse but could not be delivered due to incomplete address information,” reads one typical message.

A link in the message points to a realistic website where you are asked to enter more details and make a small payment—while behind the scenes, cybercriminals hoover up your information and credit card digits in real time.

These messages originate from one prolific collection of loosely linked cybercriminals: “smishing” syndicates.

Over the past three years, these Chinese-speaking fraudsters have developed and operated the world’s foremost smishing operation, spamming millions of people with text messages and likely stealing millions of dollars in the process. The term “smishing” is a mashup of SMS and phishing emails, which try to trick people into handing over personal details. Text messages, though, add a layer of urgency and may catch people off-guard as they go about their busy day.

Now, these groups are quickly adapting their methods and expanding their scamming, security experts say.

“They operate very similar to a \[legitimate\] business in a lot of ways,” says Grant Smith, the founder of offensive cybersecurity firm Phantom Security who last year hacked one Chinese-language group and uncovered how its internal phishing kits function.

“The vast majority of the kits I see nowadays are surprisingly well put together,” Smith says. “They are constantly developing these, constantly updating them, making them look better, making them more secure.”

Multiple Chinese-speaking smishing groups and individual actors are involved in the ongoing development of new techniques and running the large-scale fraud. Often, groups will even sell the kits they develop to less sophisticated cybercriminals to easily operate.

Ford Merrill, a security researcher at SecAlliance, part of the CSIS Security Group, has been tracking the syndicates for two years and says there are now seven “major” Chinese “phishing-as-a-service” actors. “They have been enabling global SMS-based phishing campaigns at a massive scale since early 2023 onwards,” he tells WIRED.

Criminals will create websites impersonating companies or brands—such as postal services, tax authorities, telecoms, utilities companies, and increasingly payment providers—and then send texts (either SMS, encrypted iMessage, or RCS) that entice people to enter their personal information and bank cards on the fraudulent websites. This process requires the fraudsters to register thousands of domains and use Apple iCloud accounts.

One of the most prominent of the smishing actors is often referred to as the Smishing Triad—although security researchers group Chinese-speaking threat actors and affiliates in different ways—which has impersonated organizations and brands in at least 121 countries, according to recent research by security company Silent Push.

Around 200,000 domains have been used by the group in recent years, the research says, with around 187 top-level domains—such as .top, .world, and .vip—being used. Across one recent 20-day period, there were more than 1 million page visits to scam websites used by the Smishing Triad, according to Silent Push.

Besides collecting names, emails, addresses, and bank card details, the websites also prompt people to enter one-time passwords or authentication codes that allow the criminals to add bank cards to Apple Pay or Google Wallet, allowing them to use the cards while on the other side of the world.

“They have effectively turned the modern digital wallet, like Apple Pay or Google Wallet, into the best card-cloning device we’ve ever invented,” Merrill says.

In Telegram groups linked to the cybercriminal organizations, some members share photos and videos of bank cards being added to digital wallets on iPhones and Androids. For instance, in one video, scammers allegedly show off dozens of virtual cards that they have added to phones they are using.

Merrill says the criminals may not make payments using the cards they’ve added to digital wallets straightaway, but it probably won’t take long.

“When we first started seeing this, they would wait between 60 and 90 days before actually stealing money from the cards,” he explains, adding that at first the criminals would let the cards “age” on a device in an attempt to look legitimate. “Nowadays you would be lucky if they wait seven days or even a couple days. Once they hit the card, they hit it hard and fast.”

“Security is core to the Google Wallet experience, and we work closely with card issuers to prevent fraud,” says Google communications manager Olivia O'Brien. “For example, banks notify customers when their card has been added to a new Wallet, and we provide signals to help issuers detect fraudulent behavior so they can decide whether to approve added cards.”

Apple did not respond to WIRED’s request for comment.

The giant scam ecosystem is powered in part by commercial underground scamming services. Findings from security firm Resecurity, which has tracked the Smishing Triad for more than two years, says the group has been using “bulk” SMS and message-sending services as it has expanded the number of messages it sends.

Meanwhile, as multiple security researchers have noted, the Smishing Triad group also uses its own software, called Lighthouse, to collect, manage, and store people's personal information and card details. A video of the Lighthouse software originally shared on Telegram and republished by Silent Push shows how the system collects card details.

The latest version of the software, which was updated in March this year, “targets dozens of financial brands” including PayPal, Mastercard, Visa, and Stripe, Silent Push says. In addition, the research says, Australian banking brands appear to be impersonated, indicating a potential further expansion of targets.

The smishing groups are constantly improving their own scamming software. Smith, the researcher who hacked a smishing gang, says he has seen groups operating their own development pipelines for their software and systems.

“Recently they were using some custom-made software that they had to basically replicate Jira and have tickets open for any issues with the platform and any customer complaints,” Smith says. “They would assign them to team members.”

Chinese smishing groups do not appear to be slowing down. The cybercriminals are behind huge waves of toll road text scams sweeping across the United States this year, says Shawn Loveland, the chief operating officer at Resecurity. “They’re increasing in their scale and their volume of attacks,” he says.

Loveland says there may be multiple ways to limit the effectiveness of smishing operations. Domain registrars could get better at detecting fraudulent websites, for example, and improved spam filtering on messages would help potential victims. Plus, law enforcement could target the platforms and systems they use to create accounts and send messages.

Making it harder for smishing groups to successfully operate could reduce profits and have a cooling effect on the surging criminal ecosystem, Loveland says.

“Criminals have a supply chain, and you don't have to go after all the components in the supply chain,” he says. “You can go after choke points in the supply chain.”

Smishing Triad: The Scam Group Stealing the World’s Riches

Allegedly responsible for the theft of $1.5 billion in cryptocurrency from a single exchange, North Korea’s TraderTraitor is one of the most sophisticated cybercrime groups in the world.

On February 21, the largest crypto heist ever started to unfold. Hackers gained control of a crypto wallet belonging to the world’s second-largest cryptocurrency exchange, Bybit, and stole almost $1.5 billion of digital tokens. They quickly shunted the money between dozens of cryptocurrency wallets and services to try and obscure the activity, before starting to cash the stolen funds out.

The eye-popping digital raid had all the hallmarks of being conducted by one of North Korea’s elite subgroups of hackers. While Bybit remained solvent by borrowing cryptocurrency and launched a bounty scheme to track down the stolen funds, the FBI quickly pinned the blame on the North Korean hackers known as TraderTraitor.

Before the Bybit heist, TraderTraitor had already been linked to other high-profile cryptocurrency thefts and compromises of supply chain software.

“We were waiting for the next big thing,” says Michael Barnhart, a longtime cybersecurity researcher focused on North Korea and investigator at security firm DTEX Systems. “They didn't go away. They didn’t try to stop. They were clearly plotting and planning—and they’re doing that now,” he says.

North Korea’s hackers—alongside those from China, Russia, and Iran—are consistently considered to be one of the most sophisticated and most dangerous cyber threats to Western democracies. While all of these countries engage in espionage and theft of sensitive data, North Korea’s cyber operations come with their own set of distinct goals: helping to fund the hermit kingdom’s nuclear programs. Increasingly, that means stealing cryptocurrency.

Over at least the past five years, the totalitarian regime of Kim Jong-un has deployed technically skilled IT workers to infiltrate companies around the world and earn wages that can be sent back to the motherland. In some cases, after being fired, those workers extort their former employers by threatening to release sensitive data. At the same time, North Korean hackers, as part of the broad umbrella Lazarus Group, have stolen billions in cryptocurrency from exchanges and companies around the world. TraderTraitor makes up one part of the wider Lazarus group, which is run out of the Reconnaissance General Bureau, the North Korean intelligence agency.

TraderTraitor—which is also referred to as Jade Sleet, Slow Pisces, and UNC4899 by security companies—is primarily interested in cryptocurrency.

“They use a variety of creative techniques to get into blockchain, cryptocurrency, anything that has to do with platforms, trading forums, all of those different things that are around cryptocurrency and decentralized finance,” says Sherrod DeGrippo, the director of threat intelligence strategy at Microsoft. “The Jade Sleet group \[TraderTraitor\] is one of the most sophisticated groups within that echelon,” she says.

TraderTraitor first emerged around the start of 2022, multiple cybersecurity researchers say, and is likely an offshoot of the North Korean APT38 group that hacked the SWIFT financial system and attempted to steal $1 billion from the Central Bank of Bangladesh at the start of 2016. “They walked off with very little money,” says DTEX Systems’s Barnhart. “In that moment you had a real, significant shift.”

Barnhart says North Korea realized that relying on other people—such as money mules—could make their operations less effective. Instead, they could steal cryptocurrency. Two groups emerged from that tactical shift, Barnhart says, CryptoCore and TraderTraitor. “TraderTraitor is the most sophisticated of all,” he says. “And why? Because APT38 was the A team.”

Since then, TraderTraitor has been linked to multiple large-scale cryptocurrency thefts in recent years. For instance, the March 2024 theft of $308 million from Japan-based cryptocurrency company DMM has been linked to TraderTraitor by the FBI, Department of Defense, and police in Japan.

TraderTraitor typically targets people working at Web3 firms using spear-phishing messages—most often, people working on software development. “They know the individuals that work at these companies, they track them, they have profiles on them, they know which trading platforms are doing the most volume. They’re focused on that entire industry, understanding it backwards and forwards,” says Microsoft’s DeGrippo.

GitHub, which is owned by Microsoft, highlighted in July 2023 how TraderTraitor created fake accounts on the coding platform, plus LinkedIn, Slack, and Telegram. The TraderTraitor criminals can create fake personas that they use to message their targets or use real accounts that have been hacked, GitHub’s research says. In that instance, TraderTraitor invited developers to collaborate on GitHub, before ultimately infecting them with malware using malicious code. Recently, security researchers at Palo Alto Networks’ Unit 42 threat intelligence team found 50 North Korean recruiter profiles on LinkedIn and linked them back to TraderTraitor.

The group has been seen using “custom backdoors,” such as PLOTTWIST and TIEDYE, that target macOS, says Adrian Hernandez, a senior threat analyst at Google’s Threat Intelligence Group. “These are typically heavily obfuscated to prevent detection and thwart analysis,” Hernandez says. “Once UNC4899 \[TraderTraitor\] has gained access to valid credentials, we’ve observed this threat actor moving laterally and accessing other accounts to access hosts and systems, keeping a low profile and aiming to evade detection.”

Once the North Korean hackers have their hands on cryptocurrency or digital wallets, the money laundering often follows a similar pattern, as cryptocurrency tracing firm Elliptic outlined in a blog post breaking down the Bybit hack. To avoid having cryptocurrency wallets frozen, they quickly swap stolen tokens—which are often issued by centralized entities and can have restrictions placed upon them—for more mainstream cryptocurrency assets like ether and bitcoin that are harder to limit.

“The second step of the laundering process is to ‘layer’ the stolen funds in order to attempt to conceal the transaction trail,” Elliptic writes. This means splitting the funds into smaller amounts and sending them to multiple wallets. With Bybit, Elliptic writes, money was sent to 50 different wallets that were then emptied in the coming days. This cryptocurrency is then moved through various cryptocurrency exchanges, converted into bitcoin, and passed through crypto mixers that aim to obscure crypto transactions.

“North Korea is the most sophisticated and well-resourced launderer of crypto assets in existence, continually adapting its techniques to evade identification and seizure of stolen assets,” Elliptic says in its blog post.

In addition to cryptocurrency heists, TraderTraitor has been linked to hacks at software supply chain companies, most prominently JumpCloud in June 2023. Compromising software used by multiple companies may provide the hackers a stealthier way into their intended targets. “That could impact any tech industry, any organization that uses that software,” says Andy Piazza, senior director for threat research at Unit 42.

As TraderTraitor has increasingly garnered attention over the past couple of years, Piazza says he has seen the group improve their operations and attempt to evade detection. For example, Unit 42’s recent research noted that TraderTraitor had been using malware the researchers called RN Loader that installs an information stealer and then deletes itself, making it harder to detect.

“You can definitely tell that they’re stepping up,” Piazza says.

Piazza says that unlike haphazard Russian hacking groups—which were both in the networks of the DNC simultaneously around 2016—there appears to be more organization with the North Korean groups. “It seems more coordinated that they're not bumping into each other out in the battle space,” Piazza says. “They’re really showing that they have the capability to be focused on that OPSEC, to be focused on that persistence capability.”

North Korea’s hacking operations may be even more complex than many realize. According to Piazza and other experts WIRED spoke to, the crypto hackers and the undercover IT workers may even coordinate. Their tactics show some “overlap,” Piazza says.

“If you right now went out onto some type of freelance website and said that you are a brand-new crypto startup and you’re looking for developers before the day is out, you would have North Koreans in your inbox,” Barnhart, the DTEX Systems researcher, says. He says some North Korean hackers can bounce between the country’s different groups, and there’s the possibility that they could also work with or alongside its IT workers. There may be more overlap than people thought, Barnhart says.

“Whenever we attribute this \[hacking\] back to TraderTraitor, was nobody else involved? Did somebody else have a hand in there?”

TraderTraitor: The Kings of the Crypto Heist

After a series of setbacks, the notorious Black Basta ransomware gang went underground. Researchers are bracing for its probable return in a new form.

The pecking order of ransomware gangs is always shifting and evolving, with the most aggressive and reckless groups netting big payouts from vulnerable targets—but often ultimately flaming out. Russian-speaking group Black Basta is the latest example of the trend having stalled out in recent months due to takedowns by law enforcement and a damaging leak. But after some quiet weeks, researchers warn that, far from being dead and gone, the actors involved with Black Basta will reemerge in other cybercriminal groups—or potentially already have—to start the cycle once again.

Since appearing in April 2022, Black Basta has generated hundreds of millions of dollars in payments targeting an array of corporate victims in health care, critical infrastructure, and other high-stakes industries. The group uses double extortion to pressure targets into paying a ransom—stealing data and threatening to leak it while also encrypting a target’s systems to hold them hostage. The US Cybersecurity and Infrastructure Security Agency warned last year that Black Basta had gone on a spree targeting more than 500 organizations in North America, Europe, and Australia.

A major international law enforcement takedown in 2023 of the “Qakbot” botnet hindered Black Basta’s operations, though. And, this February, a major leak of the group’s internal data—including chat logs and operational information—rocked the group. Since then, it has gone dormant. Researchers warn, though, that the criminals behind Black Basta are already on the move and are almost certain to stage a resurgence.

“We haven’t seen the leaders of Black Basta regroup, but they’re going to continue to work, they’re going to continue to operate,” says Allan Liska, a threat intelligence analyst focused on ransomware at the security firm Recorded Future. “There’s still too much money in it not to. And ransomware actors are creatures of habit just like anyone.”

The leak revealed details about Black Basta’s malware and technical capabilities, its internal squabbles, and clues about the identity of the actors behind the group, particularly its main administrator. The exposed data was from what might be considered Black Basta’s heyday, September 2023 to September 2024. During this period, the group didn’t shy away from the possibility of causing harm with its breaches. A particularly aggressive attack last year on the St. Louis–based health care network Ascension, for example, reportedly caused disruptions in care, including rerouted ambulances.

Black Basta struggled to maintain its momentum, though, after the 2023 Qakbot takedown, known as Operation Duck Hunt.

“It was a huge blow to them, and they were trying to get back on their feet—use other botnets, work on a custom botnet, but that didn’t really work, and ultimately their infection rate was declining,” says Yelisey Bohuslavskiy, chief research officer of the threat-intelligence firm RedSense. “They had fewer targets and were getting into fewer networks. They were still dangerous, but there was this feeling that there was deterioration going on.”

Even in this decline, there was evidence that Black Basta was trying to mount a resurgence. In addition to exploring new malware, the gang started focusing on compromising targets through social engineering and influence campaigns, particularly spam email operations and tech support scams. But after the leak, Bohuslavskiy says, members began moving to other groups and have already been buoying their new gangs.

Like any industry, the Russian cybercriminal landscape is full of people who have worked together or competed against one another for years. Black Basta was able to establish itself so quickly because many of its members were involved with previous cybercriminal operations, including the longtime cybercriminal gang Conti. Conti is a well-known group because of another internal leak incident in 2022 that exposed its inner workings and ties to the Kremlin. After Conti’s demise, researchers tracked its members as they dispersed and started new hacking groups, including Black Basta.

While Black Basta is not unique in its tactics and methods, researchers say that the group is noteworthy for its technical skills and depth of cybercriminal experience, which allowed it to push the envelope on the approaches a ransomware group can take.

“The people behind Black Basta have been in a lot of networks and have a lot of experience,” Recorder Future’s Liska says. “They aren’t the most prolific group but I think they are one of the more dangerous groups because they are so skilled.”

February’s leaks revealed, for example, that Black Basta developed a tool for automatically infiltrating network devices like routers that had easily-guessable passwords. Automating a tool to guess passwords is not a groundbreaking capability, but it is the type of project that many ransomware groups wouldn’t think to take on themselves or have the capacity to develop in-house.

In a report last week analyzing the leaked Black Basta communications, researchers from the security firm Trustwave wrote, “The messages show how members exhibit remarkable autonomy and creativity, adapting quickly to evolving security landscapes.”

The Black Basta leak is a cache of 200,000 messages and other data apparently taken from the group’s Matrix chat server, bestflowers247.online, by user “ExploitWhispers.” The trove includes the text of the group’s communications plus time stamps, sender and recipient details, and other metadata. The identity and motivation of “ExploitWhispers” is unknown, but they claimed to have leaked the data because Black Basta had allegedly attacked Russian banks, violating the unwritten rule that cybercriminals can operate in Russia with impunity so long as they do not attack Russian organizations.

While the exposure that came with the leaks was a death knell for Black Basta as a group, it is more likely to be a setback than a permanent defeat for its members.

“We haven’t seen the leaders of Black Basta regroup, but they’re going to continue to work, they’re going to continue to operate,” Recorded Future’s Liska says. “There’s still too much money in it not to. And ransomware actors are creatures of habit just like anyone.”

RedSense’s Bohuslavskiy adds that he has already seen signs of Black Basta members cropping up in other active gangs, including “BlackSuit,” “INC,” “Lynx,” “Cactus,” and “Nokoyawa.”

“Now that Black Basta is done, a lot of the people have migrated, and there are a number of other ransomware groups that are getting infusions of Black Basta talent,” Bohuslavskiy says.

Black Basta: The Fallen Ransomware Gang That Lives On

An email sent by the Department of Homeland Security instructs people in the US on a temporary legal status to leave the country. But who the email actually applies to—and who actually received it—is far from clear.

The United States Department of Homeland Security sent an email this week informing people living in the US on a temporary legal status that their "parole" has been revoked and instructed them to leave the country "immediately." However, the email was also addressed to at least one US citizen, an immigration attorney from Massachusetts. And it remains unclear who must abide by the email's instructions—or why the apparent revocation of legal immigration status was delivered via email at all.

The email informs the recipient that “DHS is now exercising its discretion to terminate your parole,” which it says will go into effect “7 days from the date of this notice.” The email appears to be similar, if not identical, to messages received by users of CBP One, an app developed during the Biden administration that allows non-citizens from certain countries to schedule appointments at US points of entry in an effort to seek asylum. A spokesperson for US Customs and Border Protection (CBP) tells WIRED, however, that the email was sent more broadly.

“CBP has issued notices terminating parole for individuals who do not have lawful status to remain,” says CBP assistant commissioner for public affairs Hilton Beckham. “This process is not limited to CBP One users and does not currently apply to those paroled under programs such as U4U and OAW.”

U4U refers to Uniting for Ukraine, a program launched under the Biden administration to allow for expedited immigration to the US for Ukrainians fleeing Russia’s war against its neighboring country. Former President Joe Biden began OAW, or Operation Allies Welcome, in 2021 following the US military’s exit from Afghanistan to allow for the safe resettling of “vulnerable Afghans, including those who worked alongside us in Afghanistan for the past two decades,” according to DHS.

The email itself does not identify these or any other exemptions, nor does it make clear to whom it applies beyond the recipient. Beckham also confirmed that the email was sent to whatever email address the agency had associated with the intended target, leading to confusion for at least one immigration attorney.

“Some personal news: the Department of Homeland Security has given me, an immigration lawyer born in Newton, Massachusetts, seven days to leave the US,” wrote Nicole Micheroni, a partner at  Cameron Micheroni & Silvia LLC, in a post on Bluesky on Friday night. “Does anyone know if you can get Italian citizenship through great-grandparents?”

Micheroni tells WIRED that she first thought the email was intended for one of her clients, but she quickly noticed that it was only addressed to her.

“I don't feel like I'm actually going to be deported in seven days, but it's concerning that this is the level of care they're using to send these out,” Micheroni says. She adds that it’s possible that the DHS email was “intended for one of my clients or somebody else,” as it’s not uncommon for immigrants in the US to list their attorneys as the point of contact.

The Trump administration has sought to revoke the parole of some 532,000 Cubans, Haitians, Nicaraguans, and Venezuelans who entered the US under a Biden-era humanitarian parole program. While it moved to subject them to expedited deportation effective April 24, on Thursday a federal judge in Boston said she would issue a protective order blocking that attempt. However, CBP's Beckham says people covered by the order did not receive the email, which stipulates that it does not apply to people who “have otherwise obtained a lawful basis to remain” in the US.

“Thursday’s order pertains to the CHNV program, a categorical parole that was not sent termination notices,” Beckham says. “Aliens from those countries who did not enter through the CHNV program may have been subject to the termination notices.”

Attorney Lauren Regan, founder and executive director of the nonprofit Civil Liberties Defense Center, tells WIRED that the lack of clarity about whether the revocation of temporary parole applies to the recipient of the email is likely causing fear and confusion among many immigrants, especially those without access to adequate legal guidance.

“So many people don’t have a lawyer, or their lawyer has 6,000 clients,” Regan says, which “completely overloads” the attorneys who often provide pro bono legal services to immigrants.

“A lot of people that are here on parole status don't know the nuances of immigration law, so they get this email and they don't know if it applies to them,” Micheroni says. “And most of them assume that it does because everything is really scary for people right now.”

It’s also unclear whether the email is related to recent efforts by Elon Musk’s so-called Department of Government Efficiency (DOGE). In an April 10 post on X, DOGE claimed that “CBP identified a subset of 6.3k individuals paroled into the United States since 2023 on the FBI's Terrorist Screening Center watchlist or with criminal records. These paroles have since been terminated with immediate effect.”

Beckham did not immediately respond to questions about whether the email was intended for these 6,300 individuals, nor did she answer WIRED’s questions about how many people received the email.

Then there’s the matter of the email being an email at all. Regan says that “it is absolutely not common” for a change in legal immigration status to arrive via email, which typically happens in person or via certified mail. “People would think it’s a phishing email or something not legitimate,” Regan says. Also, the fact that the email does not appear to have been first posted on a government website added to questions about its authenticity.

“Normally if the government is going to change a practice, they would first do it on their websites,” Regan says, adding, “but the fact that this was not on the website first and then sent out as a direct communication is very, very unusual.”

Regan also notes that many immigrants do not have email addresses, and therefore couldn’t receive the communication in the first place.

Even for Micheroni, a US citizen and immigration attorney, the Trump administration’s aggressive immigration enforcement practices have made life less stable. The email only made matters worse.

“I have gotten some serious inquiries from my parents or other family members or friends being like, ‘what do I do if you stop answering me or if you disappear? Like, who do you want me to call?’” she says.

“And if people in my life are feeling this way, and this is what I do, I know a lot about it,” Micheroni adds. “I can't imagine what it's like for people that don't fully understand immigration law.”

_Updated at 4 pm ET: Added additional comment from CBP clarifying that Cubans, Haitians, Nicaraguans, and Venezuelans who entered the US under the so-called CHNV program did not receive the email._

Homeland Security Email Tells a US Citizen to 'Immediately' Self-Deport

Plus: The Department of Homeland Security begins surveilling immigrants' social media, President Donald Trump targets former CISA director who refuted his claims of 2020 election fraud, and more.

The Israeli spyware maker NSO Group has been on the US Department of Commerce “blacklist” since 2021 over its business of selling targeted hacking tools. But a WIRED investigation has found that the company now appears to be working to stage a comeback in Trump's America, hiring a lobbying firm with the ties to the administration to make its case.

As the White House continues its massive gutting of the United States federal government, remote and hybrid workers have been forced back to the office in a poorly coordinated effort that has left critical employees without necessary resources—even reliable Wi-Fi. And Elon Musk’s so-called Department of Government Efficiency (DOGE) held a “hackathon” in Washington, DC, this week to work on developing a “mega API” that could act as a bridge between software systems for accessing and sharing IRS data more easily.

Meanwhile, new research this week indicates that misconfigured sexual fantasy-focused AI chatbots are leaking users' chats on the open internet—revealing explicit prompts and conversations that in some cases include descriptions of child sexual abuse.

And there's more. Each week, we round up the security and privacy news we didn’t cover in depth ourselves. Click the headlines to read the full stories, and stay safe out there.

In a secret December meeting between the US and China, Beijing officials claimed credit for a broad hacking campaign that has compromised US infrastructure and alarmed American officials, according to Wall Street Journal sources. Tensions between the two countries have escalated sharply in recent weeks, because of President Donald Trump's trade war.

In public and private meetings, Chinese officials are typically firm in their denials about any and all accusations of offensive hacking. This makes it all the more unusual that the Chinese delegation specifically confirmed that years of attacks on US water utilities, ports, and other targets are the result of the US's policy support of Taiwan. Security researchers refer to the collective activity as having been perpetrated by the actor “Volt Typhoon.”

Meanwhile, the National Counterintelligence and Security Center, along with the FBI and Pentagon’s counterintelligence service, issued an alert this week that China’s intelligence services have been working to recruit current and former US federal employees by posing as private organizations like consulting firms and think tanks to establish connections.

**DHS Is Now Monitoring Immigrants’ Social Media for Antisemitism**

US Citizenship and Immigration Services said on Wednesday that it is starting to monitor immigrants' social media activity for signs of antisemitic activity and “physical harassment of Jewish individuals.” The agency, which operates under the Department of Homeland Security, said that such behavior would be grounds for “denying immigration benefit requests.” The new policy applies to people applying for permanent residence in the US as well as students and other affiliates of “educational institutions linked to antisemitic activity.” The move comes as Immigration and Customs Enforcement has made controversial arrests of pro-Palestinian student activists, including Mahmoud Khalil of Columbia University and Rumeysa Ozturk of Tufts University, over alleged antisemitic activity. Their lawyers deny the allegations.

**Trump Revokes Security Clearance, Orders Investigation of Ex-CISA Director**

President Trump this week ordered a federal investigation into former US Cybersecurity and Infrastructure Security Agency director Chris Krebs. An executive order on Wednesday revoked Krebs’ security clearance and also directed the Department of Homeland Security and the US attorney general to conduct the review. Krebs was fired by Trump in November 2020 during his first term after Krebs publicly refuted Trump’s claims of election fraud during that year's presidential election. The executive order alleges that by debunking false claims about the election while in office, Krebs violated the First Amendment's prohibition on government interference in freedom of expression.

In addition to removing Krebs' clearance, the order also revokes the clearances of anyone who works at Krebs' current employer, the security firm SentinelOne. The company said this week in a statement that it “will actively cooperate in any review of security clearances held by any of our personnel” and emphasized that the order will not result in significant operational disruption, because the company only has a handful of employees with clearances.

**NSA, Cyber Command Officials Cancel RSA Security Conference Appearances**

NSA Cybersecurity Division Director Dave Luber and Cyber Command Executive Director Morgan Adamski will no longer speak at the prominent RSA security conference, scheduled to begin on April 28 in San Francisco. Both appeared at the conference last year. A source told Nextgov/FCW that the cancellations were the result of agency restrictions on nonessential travel. RSA typically features top US national security and cybersecurity officials alongside industry players and researchers. President Trump recently fired General Timothy Haugh, who led both the NSA and US Cyber Command.

China Secretly (and Weirdly) Admits It Hacked US Infrastructure

Some misconfigured AI chatbots are pushing people’s chats to the open web—revealing sexual prompts and conversations that include descriptions of child sexual abuse.

Several AI chatbots designed for fantasy and sexual role-playing conversations are leaking user prompts to the web in almost real time, new research seen by WIRED shows. Some of the leaked data shows people creating conversations detailing child sexual abuse, according to the research.

Conversations with generative AI chatbots are near instantaneous—you type a prompt and the AI responds. If the systems are configured improperly, however, this can lead to chats being exposed. In March, researchers at the security firm UpGuard discovered around 400 exposed AI systems while scanning the web looking for misconfigurations. Of these, 117 IP addresses are leaking prompts. The vast majority of these appeared to be test setups, while others contained generic prompts relating to educational quizzes or nonsensitive information, says Greg Pollock, director of research and insights at UpGuard. “There were a handful that stood out as very different from the others,” Pollock says.

Three of these were running-role playing scenarios where people can talk to a variety of predefined AI “characters”—for instance, one personality called Neva is described as a 21-year-old woman who lives in a college dorm room with three other women and is “shy and often looks sad.” Two of the role-playing setups were overtly sexual. “It’s basically all being used for some sort of sexually explicit role play,” Pollock says of the exposed prompts. “Some of the scenarios involve sex with children.”

Over a period of 24 hours, UpGuard collected prompts exposed by the AI systems to analyze the data and try to pin down the source of the leak. Pollock says the company collected new data every minute, amassing around 1,000 leaked prompts, including those in English, Russia, French, German, and Spanish.

It was not possible to identify which websites or services are leaking the data, Pollock says, adding it is likely from small instances of AI models being used, possibly by individuals rather than companies. No usernames or personal information of people sending prompts were included in the data, Pollock says.

Across the 952 messages gathered by UpGuard—likely just a glimpse of how the models are being used—there were 108 narratives or role-play scenarios, UpGuard’s research says. Five of these scenarios involved children, Pollock adds, including those as young as 7.

“LLMs are being used to mass-produce and then lower the barrier to entry to interacting with fantasies of child sexual abuse,” Pollock says. “There's clearly absolutely no regulation happening for this, and it seems to be a huge mismatch between the realities of how this technology is being used very actively and what the regulation would be targeted at.”

WIRED last week reported that a South Korea–based image generator was being used to create AI-generated child abuse and exposed thousands of images in an open database. The company behind the website shut the generator down after being approached by WIRED. Child-protection groups around the world say AI-generated child sexual abuse material, which is illegal in many countries, is growing quickly and making it harder to do their jobs. The UK’s anti-child-abuse charity has also called for new laws against generative AI chatbots that “simulate the offence of sexual communication with a child.”

All of the 400 exposed AI systems found by UpGuard have one thing in common: They use the open source AI framework called llama.cpp. This software allows people to relatively easily deploy open source AI models on their own systems or servers. However, if it is not set up properly, it can inadvertently expose prompts that are being sent. As companies and organizations of all sizes deploy AI, properly configuring the systems and infrastructure being used is crucial to prevent leaks.

Rapid improvements to generative AI over the past three years have led to an explosion in AI companions and systems that appear more “human.” For instance, Meta has experimented with AI characters that people can chat with on WhatsApp, Instagram, and Messenger. Generally, companion websites and apps allow people to have free-flowing conversations with AI characters—portraying characters with customizable personalities or as public figures such as celebrities.

People have found friendship and support from their conversations with AI—and not all of them encourage romantic or sexual scenarios. Perhaps unsurprisingly, though, people have fallen in love with their AI characters, and dozens of AI girlfriend and boyfriend services have popped up in recent years.

Claire Boine, a postdoctoral research fellow at the Washington University School of Law and affiliate of the Cordell Institute, says millions of people, including adults and adolescents, are using general AI companion apps. “We do know that many people develop some emotional bond with the chatbots,” says Boine, who has published research on the subject. “People being emotionally bonded with their AI companions, for instance, make them more likely to disclose personal or intimate information.”

However, Boine says, there is often a power imbalance in becoming emotionally attached to an AI created by a corporate entity. “Sometimes people engage with those chats in the first place to develop that type of relationship,” Boine says. “But then I feel like once they've developed it, they can't really opt out that easily.”

As the AI companion industry has grown, some of these services lack content moderation and other controls. Character AI, which is backed by Google, is being sued after a teenager from Florida died by suicide after allegedly becoming obsessed with one of its chatbots. (Character AI has increased its safety tools over time.) Separately, users of the generative AI tool Replika were upended when the company made changes to its personalities.

Aside from individual companions, there are also role-playing and fantasy companion services—each with thousands of personas people can speak with—that place the user as a character in a scenario. Some of these can be highly sexualized and provide NSFW chats. They can use anime characters, some of which appear young, with some sites claiming they allow “uncensored” conversations.

“We stress test these things and continue to be very surprised by what these platforms are allowed to say and do with seemingly no regulation or limitation,” says Adam Dodge, the founder of Endtab (Ending Technology-Enabled Abuse). “This is not even remotely on people’s radar yet.” Dodge says these technologies are opening up a new era of online pornography, which can in turn introduce new societal problems as the technology continues to mature and improve. “Passive users are now active participants with unprecedented control over the digital bodies and likenesses of women and girls,” he says of some sites.

While UpGuard’s Pollock could not directly connect the leaked data from the role-playing chats to a single website, he did see signs that indicated character names or scenarios could have been uploaded to multiple companion websites that allow user input. Data seen by WIRED shows that the scenarios and characters in the leaked prompts are hundreds of words long, detailed, and complex.

“This is a never-ending, text-based role-play conversation between Josh and the described characters,” one of the system prompts says. It adds that all the characters are over 18 and that, in addition to “Josh,” there are two sisters who live next door to the character. The characters’ personalities, bodies, and sexual preferences are described in the prompt. The characters should “react naturally based on their personality, relationships, and the scene” while providing “engaging responses” and “maintain a slow-burn approach during intimate moments,” the prompt says.

“When you go to those sites, there are hundreds of thousands of these characters, most of which involve pretty intense sexual situations,” Pollock says, adding the text based communication mimics online and messaging group chats. “You can write whatever sexual scenarios you want, but this is truly a new thing where you have the appearance of interacting with them in almost exactly the same way you interact with a lot of people.” In other words, they’re designed to be engaging and to encourage more conversation.

That can lead to situations where people may overshare and create risks. “If people are disclosing things they’ve never told anyone to these platforms and it leaks, that is the Everest of privacy violations,” Dodge says. “That’s an order of magnitude we've never seen before and would make really good leverage to sextort someone.”

Source

Wired