The breach of the right-wing provocateur was simply a way of “stirring up some drama,” the attacker tells WIRED. But the damage could have been much worse.

The hacker who claims to have compromised the Twitter account of right-wing commentator Matt Walsh last night says he meant no harm—though he was clearly in a position to exact some. The whole purpose of the attack, he says, was to stir up controversy and sow chaos on Twitter. 

If that was all he was after, well, mission accomplished.

Walsh’s account first appeared compromised Tuesday night after a series of out-of-character posts appeared on his feed. They included jabs at fellow conservative media figures, including colleague and Daily Wire host Ben Shapiro. In a tweet directed at Shapiro, the hacker, who goes by the alias Doomed, wrote: “You Know What You Did, You Are A Closeted Homosexual And Hide Behind Being Jewish.”

Other tweets included “Joe Rogan Is A Pedophile,” and “I Can Confirm Andrew Tate Kidnapped And R\*ped Those Girls,” a reference to the former British kickboxer and controversial misogynist arrested in Romania last year on allegations of human trafficking.

Walsh, who hosts a Daily Wire podcast named after himself, has been frequently characterized as an “online troll.” He’s previously described himself as a “theocratic fascist,” claims to believe anime is “satanic,” and pushed for state executions of doctors who provide minors with gender-affirming care.

Due to his efforts to foment attacks against the LGBTQ community online and in media, denigrating trans people in particular as “groomers,” Walsh drew criticism more recently over a series of resurfaced remarks in which he refers to girls as young as 16 as being the “most fertile.”

“The intent was to make funny tweets, as Matt Walsh likes to ‘trigger’ people,” said Doomed, who declined to provide his real name. “We caused no financial harm, threatened anyone, \[nor\] ruined anything.” It was merely, he says, “a few silly words on social media.”

A pinned tweet on Walsh’s profile, which was suspended a few hours after the hacks began, read: “My Pronouns Are That/N\*\*\*a.”

The hack was accomplished, Doomed says, using a technique known as SIM swapping. The attack typically involves hackers tricking a cellular provider into switching a victim’s phone number to a SIM card the hackers control, rather than the one in the victim’s phone. Doomed, however, claims Walsh’s phone was compromised with the help of an “insider.”

In addition to his Twitter account, Walsh’s Google and Microsoft accounts appeared to have been compromised, granting Doomed unfettered access to the right-wing host’s private emails.

Several screenshots were provided as proof of the intrusion, including an apparent copy of Walsh’s W2 tax form, which lists his employer as Bentkey Services, LLC, the publisher of the Daily Wire. (Business records list Shapiro as the company’s director.) Other images included a direct message on Twitter from Shapiro from 2017; emails between Walsh and the conservative commentator Steven Crowder, host of the _Louder with Crowder_ podcast, dated March 2014; and a photo of Walsh holding up a paper dated November 2020—apparently used to authenticate some type of accompanying message.

The Hacker Who Hijacked Matt Walsh’s Twitter Was Just ‘Bored’

Generative AI is a tool, which means it can be used by cybercriminals, too. Here’s how to protect yourself.

The AI landscape has started to move very, very fast: consumer-facing tools such as Midjourney and ChatGPT are now able to produce incredible image and text results in seconds based on natural language prompts, and we're seeing them get deployed everywhere from web search to children's books.

However, these AI applications are being turned to more nefarious uses, including spreading malware. Take the traditional scam email, for example: It's usually littered with obvious mistakes in its grammar and spelling—mistakes that the latest group of AI models don't make, as noted in a recent advisory report from Europol.

Think about it: A lot of phishing attacks and other security threats rely on social engineering, duping users into revealing passwords, financial information, or other sensitive data. The persuasive, authentic-sounding text required for these scams can now be pumped out quite easily, with no human effort required, and endlessly tweaked and refined for specific audiences.

In the case of ChatGPT, it's important to note first that developer OpenAI has built safeguards into it. Ask it to "write malware" or a "phishing email" and  it will tell you that it's "programmed to follow strict ethical guidelines that prohibit me from engaging in any malicious activities, including writing or assisting with the creation of malware."

ChatGPT won't code malware for you, but it's polite about it.

OpenAI via David Nield

However, these protections aren't too difficult to get around: ChatGPT can certainly code, and it can certainly compose emails. Even if it doesn't know it's writing malware, it can be prompted into producing something like it. There are already signs that cybercriminals are working to get around the safety measures that have been put in place.

We're not particularly picking on ChatGPT here, but pointing out what's possible once large language models (LLMs) like it are used for more sinister purposes. Indeed, it's not too difficult to imagine criminal organizations developing their own LLMs and similar tools in order to make their scams sound more convincing. And it's not just text either: Audio and video are more difficult to fake, but it's happening as well.

When it comes to your boss asking for a report urgently, or company tech support telling you to install a security patch, or your bank informing you there's a problem you need to respond to—all these potential scams rely on building up trust and sounding genuine, and that's something AI bots are doing very well at. They can produce text, audio, and video that sounds natural and tailored to specific audiences, and they can do it quickly and constantly on demand.

So is there any hope for us mere humans in the wave of these AI-powered threats? Is the only option to give up and accept our fate? Not quite. There are still ways you can minimize your chances of getting scammed by the latest technology, and they aren't so different from the precautions you should already be thinking about.

How to Guard Against AI-powered Scams

There are two types of AI-related security threats to think about. The first involves tools such as ChatGPT or Midjourney being used to get you to install something you shouldn't, like a browser plugin. You could be tricked into paying for a service when you don't need to, perhaps, or using a tool that looks official but isn't.

To avoid falling into these traps, make sure you're up to date with what's happening with AI services like the ones we've mentioned, and always go to the original source first. In the case of ChatGPT for example, there's no officially approved mobile app, and the tool is web-only. The standard rules apply when working with these apps and their spinoffs: Check their history, the reviews associated with them, and the companies behind them, just as you would when installing any new piece of software.

The second type of threat is potentially more dangerous: AI that’s used to create text, audio, or video that sounds convincingly real. The output might even be used to mimic someone you know—like the case of the voice recording purportedly from a chief executive asking for an urgent release of funds, which duped a company employee.

While the technology may have evolved, the same techniques are still being used to try and get you to do something urgently that feels slightly (or very) unusual. Take your time, double-check wherever possible using different methods (a phone call to check an email or vice versa), and watch out for red flags—a time limit on what you're being asked to do, or a task that's out of the ordinary.

As always, keep your software and systems up to date.

Microsoft via David Nield

Following links you're not expecting from texts and emails is usually not a good idea, especially when you're being asked to log in somewhere. If your bank has apparently got in touch with a message, for example, go to the bank website directly in your browser to log in, rather than following any embedded link.

Keeping your operating systems, apps, and browsers up to date is a must (and this mostly happens automatically now, so there's no excuse). The most recent browsers will protect you against a whole host of phishing and scam attacks, whether the prompt designed to dupe you has been generated by AI or not.

There's no foolproof tool for detecting the presence of AI text, audio, or video at the moment, but there are certain signs to look out for: Think blurring and inconsistencies in pictures, or text that sounds generic and vague. While scammers may have scraped details about your life or your workplace from somewhere, it's unlikely that they know all the ins and outs of your operations.

In short, be cautious and question everything—that was true before the dawn of these new AI services, and it's true now. Like the face-morphing masks of the Mission: Impossible film series (which remain science fiction for now), you need to be absolutely sure that you're dealing with who you think you're dealing with before revealing anything.

How ChatGPT—and Bots Like It—Can Spread Malware

More than half of the enterprise routers researchers bought secondhand hadn’t been wiped, exposing sensitive info like login credentials and customer data.

You know that you're supposed to wipe your smartphone or laptop before you resell it or give it to your cousin. After all, there's a lot of valuable personal data on there that should stay in your control. Businesses and other institutions need to take the same approach, deleting their information from PCs, servers, and network equipment so it doesn't fall into the wrong hands. At the RSA security conference in San Francisco next week, though, researchers from the security firm ESET will present findings showing that more than half of secondhand enterprise routers they bought for testing had been left completely intact by their previous owners. And the devices were brimming with network information, credentials, and confidential data about the institutions they had belonged to.

The researchers bought 18 used routers in different models made by three mainstream vendors: Cisco, Fortinet, and Juniper Networks. Of those, nine were just as their owners had left them and fully accessible, while only five had been properly wiped. Two were encrypted, one was dead, and one was a mirror copy of another device. 

All nine of the unprotected devices contained credentials for the organization's VPN, credentials for another secure network communication service, or hashed root administrator passwords. And all of them included enough identifying data to determine who the previous owner or operator of the router had been. 

Eight of the nine unprotected devices included router-to-router authentication keys and information about how the router connected to specific applications used by the previous owner. Four devices exposed credentials for connecting to the networks of other organizations—like trusted partners, collaborators, or other third parties. Three contained information about how an entity could connect as a third party to the previous owner's network. And two directly contained customer data.

“A core router touches everything in the organization, so I know all about the applications and the character of the organization—it makes it very, very easy to impersonate the organization,” says Cameron Camp, the ESET security researcher who led the project. “In one case, this large group had privileged information about one of the very large accounting firms and a direct peering relationship with them. And that’s where to me it starts to get really scary, because we’re researchers, we’re here to help, but where are the rest of those routers?" 

The big danger is that the wealth of information on the devices would be valuable to cybercriminals and even state-backed hackers. Corporate application logins, network credentials, and encryption keys have high value on dark web markets and criminal forums. Attackers can also sell information about individuals for use in identity theft and other scamming. 

Details about how a corporate network operates and the digital structure of an organization are also extremely valuable, whether you're doing reconnaissance to launch a ransomware attack or plotting an espionage campaign. For example, routers may reveal that a particular organization is running outdated versions of applications or operating systems that contain exploitable vulnerabilities, essentially giving hackers a road map of possible attack strategies. And the researchers even found details on some routers about the physical building security of the previous owners' offices.

Since secondhand equipment is discounted, it would potentially be feasible for cybercriminals to invest in purchasing used devices to mine them for information and network access and then use the information themselves or resell it. The ESET researchers say that they debated whether to release their findings, because they didn't want to give cybercriminals new ideas, but they concluded that raising awareness about the issue is more pressing. 

“One of the big concerns I have is that, if somebody evil _isn’t_ doing this, it's almost hacker malpractice, because it would be so easy and obvious,” Camp says.

Eighteen routers is a tiny sample out of the millions of enterprise networking devices circulating around the world on the resale market, but other researchers say they've repeatedly seen the same issues in their work as well.

“We’ve purchased all sorts of embedded devices online on eBay and other secondhand sellers, and we’ve seen a lot that have not been digitally wiped,” says Wyatt Ford, engineering manager at Red Balloon Security, an internet-of-things security firm. “These devices can contain troves of information that can be used by bad actors in targeting and carrying out attacks.”

As in the ESET findings, Ford says that Red Balloon researchers have found passwords and other credentials and personally identifying information. Some data like usernames and configuration files are usually in plaintext and easily accessible, while passwords and configuration files are often protected because they are stored as scrambled cryptographic hashes. But Ford points out that even hashed data is still potentially at risk.  

“We’ve taken password hashes found on a device and cracked them offline—you’d be surprised how many people still base their passwords off their cats,” he says. “And even things that seem innocuous like source code, commit history, network configurations, routing rules, et cetera—they can be used to learn more about an organization, its people, and its network topology.”

The ESET researchers point out that organizations may think they're being responsible by contracting with outside device-management firms. e-waste disposal companies, or even device-sanitization services that claim to wipe big batches of enterprise devices for resale. But in practice, these third parties may not be doing what they claim. And Camp also notes that more organizations could take advantage of encryption and other security features that are already offered by mainstream routers to mitigate the fallout if devices that haven't been wiped end up loose in the world.

Camp and his colleagues tried to contact the old owners of the used routers they bought to warn them that their devices were now out in the wild spewing their data. Some were grateful for the information, but others seemed to ignore the warnings or offered no mechanism through which researchers could report security findings.

“We used trusted channels that we had to some companies, but then we found a lot of other companies are far more difficult to get a hold of,” Camp says. “Frighteningly so.”

Used Routers Often Come Loaded With Corporate Secrets

Three criminal cases detail China's alleged attempts to extend its security forces' influence online—and around the globe.

Gone are the days when China was content to merely control speech within its borders and push dissents into exile outside that strict firewall. A new set of US federal charges against dozens of Chinese police officers announced today details how the security forces of Chinese president Xi Jinping’s administration now seek to exert influence and repress critics wherever they are based around the world—with no limits to that reach, digital or physical.

This afternoon, US prosecutors in New York announced charges against 40 Chinese individuals in three different cases, each of which revolves around the alleged efforts of China’s Ministry of Public Security (MPS) to intimidate, harass, and censor critics of China overseas. 

The building in New York’s Chinatown where China’s Ministry of Public Security allegedly ran a secret police station.

Courtesy of Department of Justice

In perhaps the most shocking of the three cases, two New York-based MPS officers were arrested in New York and accused of running a secret police station in Manhattan’s Chinatown neighborhood. The hidden operation, US officials say, was devoted to hunting down and intimidating Chinese dissidents. In another case, the prosecutors announced new charges against a group of MPS officers and staff of China’s “cyberspace administration,” alleging that they had infiltrated the operations of a US tech company to disrupt videoconferences held by Chinese dissidents abroad. And in a third case, 34 MPS officers were charged with running a massive online troll farm that flooded social media platforms with fake accounts spewing disinformation and harassment.

“The trio of cases announced today details how \[China\] through its MPS has engaged in a multi-front campaign to extend the reach and impact of its authoritarian system into the United States and elsewhere around the world,” David Newman, a prosecutor in the Department of Justice’s National Security Division, told reporters at a Brooklyn press conference today. “It shows the PRC’s efforts to globalize the oppressive tactics used domestically in China to silence dissent.”

Photos of members of the Ministry of Public Security’s troll farm known as the 912 Special Projects Working Group, according to the FBI.Courtesy of FBI

Of the 40 defendants in the sprawling indictments, only the two New York-based MPS officers, Lu Jianwang and Chen Jinping, have been arrested. The two men are accused of running a front for the MPS in an unassuming building in downtown New York City. “This nondescript building in the heart of bustling Chinatown in lower Manhattan has a dark secret: Until several months ago, an entire floor of this building hosted an undeclared police station of the Chinese national police,” US Attorney Breon Peace said in the press conference.

Prosecutors describe how Lu allegedly participated in pro-China protests, harassed a Chinese national living in the US, and aided in the Chinese government’s investigation of a Chinese pro-democracy activist living in California. “In other words, the Chinese national police appear to have been using this station to track a US resident on US soil,” Peace said. He added that Lu and Chen are also accused of obstructing justice: They allegedly destroyed evidence of their communications with the Chinese government when they learned of an FBI investigation targeting them.

As surprising as it may be to learn that the Chinese government ran a secret police station in America’s largest city, it’s far from the only such operation, says Laura Harth, an activist at the human rights-focused group Safeguard Defenders. Last year, the group published a report on more than 100 such clandestine Chinese police outposts in over 50 countries, many used to pursue Chinese dissidents and critics of the Xi regime abroad.

Harth says that today’s charges against Chinese police in the US are the first of their kind. “Other countries are still pretending this is not an issue,” she says. “We’re quite happy to see this happening, both for people in the US but \[also\] for the signal it sends to other authorities around the world.”

As for the other two criminal cases announced today that target Chinese trolling, disinformation, and censorship, Harth says Safeguard Defenders hasn’t seen evidence that such troll farms are operated from within the MPS’s secret facilities overseas, but she’s not surprised to learn that they’re linked to the MPS. She says that her organization’s public communications are frequently flooded with criticisms from shady accounts that she’s long suspected were organized by the Chinese state. “It’s very tellingly troll or bot sort of work,” she says.

Prosecutors say that the troll farm of 34 MPS officers—based inside China, though prosecutors didn’t reveal where—is part of a group known as the 912 Special Project Working Group, which they say is devoted to online harassment and disinformation. According to the US Department of Justice, the 912 group posted pro-Chinese disinformation and bombarded critics of China with intimidating messages on Twitter and other platforms. The subjects of their disinformation and provocations included everything from the origins of Covid-19 to the death of George Floyd, to criticisms of the United States’ support of Ukraine in its war with Russia. In other instances, they flooded video conferences held by people critical of China with threatening and intimidating messages. And in one case, they even “drowned out the meeting with loud music and vulgar screams and threats,” according to the DOJ. 

In a separate indictment that was previously filed in 2020 but amended today, prosecutors accuse Xinjiang “Julien” Jin of working as an agent of the Chinese government inside a US telecommunications company that they refer to only as “Company-1.”

Jin, along with nine other agents of the MPS and the Cyberspace Administration of China, allegedly worked to censor Chinese dissident users of Company-1’s platform. Jin and his codefendants—nine of whom are based in China and one in Indonesia—are accused of trying to terminate meetings started by users who were critical of China, block their accounts by faking terms-of-service violations, and put them on a “quarantine” server with slower response times. 

“In the virtual sense, the threat from the PRC manifested as its agents and operators seeking to interfere with online gatherings, online organizing, and other social media critiques of the PRC regime, going so far as to conspire with an employee of a US-based technology company to compromise the privacy of online virtual meetings used as a modern means of public protest and the exercise of freedom of assembly,” Newman, the US prosecutor, told reporters today. The case, he argued, “underscores the insider threat risk facing any US company that does business in China and has employees based in China.”

Together, the three cases illustrate China’s growing desire to control dissent not just domestically, but worldwide, and to project its influence on not only the internet but the physical streets of New York. As it does so, that ambition seems bound to trigger a rare, high-stakes face-off between police agencies on opposite sides of the globe. “The efforts of the government of the PRC to export authoritarian methods to stifle free expression in the US,” said Newman, “is a threat to America’s democracy that we will not abide.”

_Update 6:30 pm ET, April 17, 2023: Added additional details about the 912 Special Project Working Group's alleged activities._

Chinese Cops Ran Troll Farm and Secret NY Police Station, US Says

The discovery of malicious encryptors for Apple computers could herald new risks for macOS users if the malware continues to evolve.

Security researchers are examining newly discovered Mac ransomware samples from the notorious gang LockBit, marking the first known example of a prominent ransomware group toying with macOS versions of its malware.

Ransomware is a pervasive threat, but attackers typically don't bother creating versions of their malware to target Macs. That's because Apple's computers, while popular, are much less prevalent than those running Windows, Linux, and other operating systems. Over the years, though, samples of seemingly experimental Mac ransomware have cropped up a couple of times, creating a sense that the risk could escalate at any moment. 

Spotted by MalwareHunterTeam, the samples of ransomware encryptors seem to have first cropped up in the malware analysis repository VirusTotal in November and December 2022, but went unnoticed until yesterday. LockBit seems to have created both a version of the encryptor targeting newer Macs running Apple processors and older Macs that ran on Apple's PowerPC chips.

Researchers say the LockBit Mac ransomware appears to be more of a first foray than anything that's fully functional and ready to be used. But the tinkering could indicate future plans, especially given that more businesses and institutions have been incorporating Macs, which could make it more appealing for ransomware attackers to invest time and resources so they can target Apple computers.

“It’s unsurprising but concerning that a large and successful ransomware group has now set their sights on macOS,” says longtime Mac security researcher and Objective-See Foundation founder Patrick Wardle. “It would be naive to assume that LockBit won’t improve and iterate on this ransomware, potentially creating a more effective and destructive version."

Apple declined to comment on the findings.

LockBit is a Russia-based ransomware gang that emerged at the end of 2019. The group is most known for its sheer volume of attacks, and for appearing well-organized and being less ostentatious and sophomoric than some of its peers in the cybercriminal landscape. But LockBit isn't immune from arrogance and public aggression. Notably, it called significant attention to itself in recent months by targeting the United Kingdom's Royal Mail and a Canadian children's hospital. 

For now, Wardle notes that LockBit's macOS encryptors seem to be in a very early phase and still have fundamental development issues like crashing on launch. And to create truly effective attack tools, LockBit will need to figure out how to circumvent macOS protections, including validity checks that Apple has added in recent years for running new software on Macs.

“In some sense, Apple is ahead of the threat, as recent versions of macOS ship with a myriad of built-in security mechanisms aimed to directly thwart, or at least reduce the impact of, ransomware attacks,” Wardle says. “However, well-funded ransomware groups will continue to evolve their malicious creations.”

Developing Mac ransomware may not be the highest priority on every attacker's to-do list, but the field is shifting. As law enforcement worldwide pushes to counter attacks, and victims increasingly have input and resources available to avoid paying, ransomware gangs are getting more desperate for new or refined strategies that will help them get paid. 

“The LockBit encryptor doesn’t look particularly viable in its current form, but I’m definitely going to be keeping an eye on it,” says Thomas Reed, director of Mac and mobile platforms at the antivirus maker Malwarebytes. “The viability may improve in the future. Or it may not, if their tests aren’t promising.”

Still, for ransomware actors looking to generate as much revenue as possible, Macs are a potentially appealing untilled field.

Apple's Macs Have Long Escaped Ransomware. That May Be Changing

Documents obtained by WIRED detail hundreds of investigations by the US agency into alleged database misuse that includes harassment, stalking, and more.

Law enforcement abuse of databases is far from unique to ICE agents. In the past decade, local police around the US have repeatedly abused their access to confidential databases. In 2016, an Associated Press investigation found that police officers across the country misused confidential law enforcement databases to get information on romantic partners, business associates, neighbors, and journalists. 

The misconduct records that WIRED obtained detail similar allegations. However, due to ICE’s sprawling access to data sets from federal, state, local, and private entities, experts are particularly concerned about how an agency with a prolific history of misconduct could abuse these tools.

Last May, Tucker and three colleagues coauthored a report called “American Dragnet: Data-Driven Deportation in the 21st Century.” Their report, based on a review of ICE spending transactions, found that the agency has amassed an enormous trove of databases containing billions of data points that enable the agency “to pull detailed dossiers on nearly anyone, seemingly at any time.”

“The databases ICE employees can access contain almost everything you might want to find about someone: who they are, where they live, where they drive, and who their family is,” says Nina Wang, a policy associate at the Center on Privacy & Technology and one of the “American Dragnet” coauthors. “All of that access to bulk data leaves the door wide open for misconduct.” 

ICE is one of 22 agencies housed under DHS. In 2021, the American Immigration Council used DHS privacy and compliance documents to compile a list of databases and information systems that DHS uses in relation to immigration law enforcement. The list captures the breadth of databases that ICE officials can access, some of which are referenced in the misconduct records. 

Among the databases that ICE agents allegedly misused were those containing medical records, license plate reader data, and biometric data. One of the most widely misused databases in the records is the Investigative Case Management (ICM) system, software developed by the data-mining firm Palantir that serves as the primary database for information collected by ICE during criminal and civil investigations.  

According to documents obtained by the Intercept, ICM allows ICE agents to access a kaleidoscope of data that reportedly includes information about “a subject’s schooling, family relationships, employment information, phone records, immigration history, foreign exchange program status, personal connections, biometric traits, criminal records, and home and work addresses.” 

Representatives from Palantir declined to be interviewed for this story. But in an email, Courtney Bowman, the company’s director of Privacy and Civil Liberties, pointed WIRED to public documents detailing ICM’s oversight mechanisms, including its ability to log and record every query an individual makes.

Much of the ICM misconduct that internal investigators flagged involves employees looking up information about themselves—so-called self-queries. While this kind of misconduct might seem harmless, Adam Schwartz, a senior staff attorney at the Electronic Frontier Foundation, asserts that it speaks to a culture of impunity at the ICE. “When you have officers taking a database intended for a specific purpose and subverting for self-gain, it exposes a lawless and dangerous mindset,” Schwartz says.

ICE Records Reveal How Agents Abuse Access to Secret Data

Plus: Hackers claim to have stolen 10 TB from Western Digital, a new spyware has emerged, and WhatsApp gets a fresh security feature.

If you had “leaking classified US military documents for the lulz” on your 2023 Bingo card, congratulations. The fast-paced drama surrounding the online disclosure of top-secret material ripped through this week’s news. We’ll dive into the details below, but there’s one key takeaway: This bizarre kind of leak may be only the beginning.

Anyone worried about chaos agents of a different variety now have a new way to protect their online identities. LinkedIn this week began to roll out new tools that allow you to verify your identity and your job. And for iOS users who want a built-in way to protect their security, we detailed how to use Apple’s all-in-one password manager.

While your personal security might be moving in the right direction, ChatGPT and other large language models (LLMs) aren’t so lucky. This week we explored the world of “jailbreaking” generative AI tools, which allows users to trick the powerful chatbots into doing things that their creators have tried to stop. It’s still early days in the world of LLM hacking, but it’s a safe bet that we’ll be hearing a lot more about this in the months to come. 

Finally, yesterday afternoon, Montana lawmakers voted to ban downloads of TikTok in the state. Governor Greg Gianforte is expected to sign the unprecedented legislation into law, despite the likelihood of swift legal and technical challenges.

But that’s not all. Each week, we round up the stories we didn’t report in-depth ourselves. Click on the headlines to read the full stories. And stay safe out there.

Yesterday, Jack Teixeira, an airman with the Massachusetts Air National Guard, was formally charged with retaining and transmitting national defense information and intentionally withholding classified documents. _The New York Times_ was the first to identify Teixeira, and FBI agents arrested him at his family’s home soon after the publication named him as the alleged source of the leak.

The 21-year-old allegedly disclosed the top-secret intelligence in an attempt to show off to members of Thug Shaker Central, an invite-only chat room hosted on Discord. The chat consisted of two dozen adolescent boys and young men who, according to _The Washington Post_, largely did not understand the significance of the leak. 

The bizarre circumstances of the leak have, according to NBC News, frustrated and embarrassed US intelligence officials who are now looking to potentially broaden the surveillance of online chat rooms after failing to spot the classified Pentagon documents that had been circulating online for weeks. 

The documents consisted of sensitive information regarding Russian military tactics during the conflict in Ukraine as well as intelligence reports on friendly nations such as Israel and South Korea, among other topics. 

Monitoring chatter in public chat rooms is commonplace for law enforcement. But if US intelligence agencies plan on surveilling private conversations without probable cause, they will run into serious legal and civil liberties hurdles, experts say. “We do not have nor do we want a system where the United States government monitors private internet chats,” Glenn Gerstell, a former general counsel of the National Security Agency, told NBC News. 

US defense secretary Lloyd Austin on Thursday said he was considering “additional measures necessary to safeguard our nation’s secrets,” and he ordered a review of “our intelligence access, accountability, and control procedures within the department to inform our efforts to prevent this kind of incident from happening again.”

Hackers who claim to have breached data storage company Western Digital earlier this month say they are holding 10 terabytes of stolen data hostage and are ready to publish it unless the company pays a “minimum 8 figure” ransom, TechCrunch reports. 

An individual who says they carried out the hack spoke to TechCrunch on Thursday, claiming to have reams of customer information. While the hacker showed TechCrunch screenshots of internal emails and contact information of Western Digital’s employees, it’s still unclear exactly what data has been stolen.

“Cut the crap, get the money, and let’s both go our separate ways,” the hackers wrote in an email to several company executives. “Simply put, let us put our egos aside and work to find a resolution to this chaotic scenario.” 

A secretive Israeli spyware company’s hacking tools have been used to target politicians and journalists in at least 10 countries, according to research by Microsoft and the University of Toronto’s Citizen Lab made public Tuesday. 

The company, QuaDream, is a small, low-profile Israeli firm that develops smartphone hacking tools intended for government clients. The firm was established in 2016 by former employees of NSO Group, the maker of the Pegasus spyware.

The QuaDream spyware targeted older versions of Apple’s iOS phone software, and it worked by sending malicious calendar invites that would not be seen by the targets, researchers say.

According to the report, Citizen Lab has located QuaDream servers in Bulgaria, the Czech Republic, Hungary, Ghana, Israel, Mexico, Romania, Singapore, the United Arab Emirates, and Uzbekistan. 

WhatsApp has introduced a new security feature that makes it harder for scammers to steal users' accounts. The feature will require individuals who download WhatsApp to a new device to use their old device to confirm their account. It's an extra layer of security that aims to protect users from account takeovers through SIM jacking or other social engineering attacks.

A WhatsApp spokesperson told Engadget that the Account Protect feature will activate only when the company suspects a malicious account takeover. If a user lost their old device, they can also request a one-time passcode from WhatsApp.

Security Roundup: Leak of Top-Secret US Intel Risks a New Wave of Mass Surveillance

If you’re worried that one of Apple’s trackers is following you without consent, try these tips.

When the AirTag launched in 2021, Apple’s Bluetooth tracker with ultra-wideband was lauded as a step toward the future of augmented reality and a great way to find everyday objects, like your lost TV remote. Cybersecurity experts expressed concern that the tracking device would be exploited by stalkers.

The warnings were prescient; multiple women reported frightening encounters where AirTags were used as stalking devices that could be slipped in a purse or taped to a car. Police departments across the United States issued warnings about the potential criminal uses of AirTags. Newer AirPods have tracking abilities similar to AirTags, but the higher cost of Apple’s earbuds limits their disposability as a tracking device.

Apple released firmware updates late in 2022 in an effort to curb misuse. Even though Tile and other competitors to the AirTag exist, the vastness of Apple’s ecosystem sets the device apart. From the US Drug Enforcement Administration using it to track international drug shipments to a man in Texas using it to find his stolen car and kill the suspect, AirTags are everywhere.

If you are concerned that a secret AirTag may be recording your location, these signs may help detect the tracker.

Signs an AirTag Is Tracking You

The type of smartphone you own affects how easily you can discover hidden AirTags. Owners of iPhones running iOS 14.5 or newer should receive a push alert whenever an unknown AirTag is nearby for an extended period of time and away from its owner. Apple’s website does not provide an exact time frame for when this alert is triggered.

Owners of newer iPhones should turn on Bluetooth and check their settings to ensure they’ll receive notifications. Under **Settings**, go to **Privacy & Security**, and toggle **Location Services** on. Scroll to the bottom of that page, tap on **System Services**, and activate **Find My iPhone**. Also, search for the **Find My** app, visit **Me** in the bottom right corner, then tap **Customize Tracking Notifications** to double-check that notifications are enabled.

When you click on the iPhone alert for an unrecognized AirTag, you may be given the option to play a sound on the AirTag to help locate it. If your iPhone runs iOS 16.2 or later, you might be able to use precision location data to find the hidden device.

Months after the release of the AirTag, Apple launched the Tracker Detect app for Android phones. Unlike the security features available for the iPhone, the Android app does not automatically look for unknown AirTags. Users must initiate the scan.

According to Eva Galperin, director of cybersecurity at the Electronic Frontier Foundation, the reason for the app’s limited functionality is complicated. “This is actually a limitation of how the Android ecosystem works and how Android apps can work,” she says. “I have called on Apple and Android to work together to incorporate the level of mitigations that Apple provides in iOS into the Android operating system, but this requires a lot of cooperation between two groups who are normally rivals.”

While some guides to finding AirTags recommend using Bluetooth scanners, Galperin does not consider this method to be reliable for tracker searching. “I have tried using various Bluetooth scanners in order to detect AirTags, and they do not work all the time,” she says.

Millions of Americans still do not own a smartphone. Without a device on hand, you must rely on visual and audible clues to find any hidden AirTags. The circular white disc is slightly larger than a quarter. As reported by _The New York Times_, Ashley Estrada discovered an AirTag lodged under her license plate, and her video documenting the incident was viewed over 20 million times on TikTok.

When the AirTag was first released, the tracker would emit a beeping noise if away from the owner for longer than three days. Apple has since shortened the time to 24 hours or less. Despite the update, you might not want to rely only on sound to detect AirTags. Numerous videos on YouTube offer DIY instructions to disable the speaker, and noiseless versions of the trackers were even listed for a short time on Etsy.

What if I Find One?

The best way to disable an AirTag is to remove the battery. To do this, flip the AirTag so the metallic side with an Apple logo is facing you. Press down on the logo and turn counterclockwise. Now you will be able to remove the cover and pop out that battery.

Apple’s support page for the AirTag suggests reaching out to the police if you believe you are in a dangerous situation. “If you feel your safety is at risk, contact your local law enforcement, who can work with Apple to request information related to the item,” the support page reads. “You might need to provide the AirTag, AirPods, Find My network accessory, and the device's serial number.” One way to figure out the serial number is to hold the top of an iPhone or other near-field-communication-enabled smartphone to the white side of an AirTag. A website with the serial number will pop up.

This page may also include a partial phone number from the person who owns the tracking device. If you feel hesitant about scanning the AirTag or do not have the ability, a serial number is printed on the device beneath the battery.

Who Does This Affect?

In the viral stories shared online and in police reports, women are often the victims of AirTag stalking, but Galperin cautions against framing unwanted tracking as solely an issue for women. “I have been working with victims of tech-enabled abuse for many years,” she says, “and I would say that about two-thirds of the survivors that come to me are women. But a third of them are men. I suspect that number would be higher if there wasn't such a stigma around being an abuse victim or survivor.”

She emphasized how men, women, and nonbinary people can all be victims of abuse, as well as perpetrators. “When we paint it all with this really broad brush, we make it really hard for victims who don’t fit that mold to come forward,” says Galperin.

For more resources, you can visit the website for the National Domestic Violence Hotline. Contact the hotline by calling 1-800-799-7233 or texting “START” to 88788.

Are You Being Tracked by an AirTag? Here’s How to Check

The state is poised to be the first in the US to block downloads of the popular app, which could ignite a precarious chain reaction for digital rights.

Montana lawmakers voted 54-43 today to ban TikTok from operating in the state and forbid app stores from offering it for download. The legislation is likely to become law, which would make Montana the first state in the US to ban the popular social media platform—a move that could spark a constitutional battle and endanger digital rights.

People who already have TikTok on their devices would not be in violation of the law, which will now go to Greg Gianforte, Montana's Republican governor. The move comes after years of amorphous assertions from the United States government under two presidential administrations that TikTok, which has 150 million US users, is a threat to national security because its parent, ByteDance, is a Chinese company.

Gianforte is expected to sign the new bill into law, which would take effect on January 1, 2024. In December, he banned TikTok from Montana government devices, a step other states have taken in recent months as well. In announcing that ban, Gianforte said, “I also encourage Montanans to protect their personal data and stop using TikTok.”

A statewide ban is radically different from a government device embargo and general encouragement, though. It has implications for Montana residents’ speech and ability to hear speech—rights protected under the US First Amendment. 

“We’re under no illusions that this is not going to get challenged,_”_ Montana attorney general Austin Knudsen told _The New York Times_ on Wednesday. "I think this is the next frontier in First Amendment jurisprudence that’s probably going to have to come from the US Supreme Court. And I think that’s probably where this is headed."

Soon after today's vote, TikTok condemned the bill on both First Amendment and logistical grounds.

“The bill's champions have admitted that they have no feasible plan for operationalizing this attempt to censor American voices and that the bill's constitutionality will be decided by the courts,” TikTok spokesperson Brooke Oberwetter said in a statement. “We will continue to fight for TikTok users and creators in Montana whose livelihoods and First Amendment rights are threatened by this egregious government overreach.”

A previous version of the bill would have required internet service providers to block connections to TikTok in Montana, a task that ISP representatives said wasn't doable. A trade association that represents companies that run mobile app stores, namely Google and Apple, also told the Montana legislature that it would be virtually impossible to halt downloads of TikTok in Montana. 

Google declined to comment. Apple did not immediately return WIRED's request for comment.

Riana Pfefferkorn, a research scholar at the Stanford Internet Observatory, says Montana attorney general Knudsen's assertions about a “next frontier in First Amendment jurisprudence” are overblown, particularly given the AG's comments during the recent _Times_ interview. In it, Knudsen specifically noted that his office was motivated to pursue a full TikTok ban after hearing protests from parents that TikTok posts included discussions of drug use, porn, and suicide.

“It's clearly unconstitutional,” Pfefferkorn says. “He admitted the purpose was to keep people from both saying and hearing legal speech. It’s an easy case."

More broadly, researchers have long warned that banning TikTok is incompatible with the democratic principles of the open internet. The US has consistently condemned platform blocks, content filtering, and internet shutdowns when other governments impose them on their citizens. These repressive digital tactics have been on the rise globally in recent years. Yet many US officials and lawmakers at both the state and federal levels have called for a TikTok ban.

“When ISPs and browser makers and tech companies are forced to develop technical methods to block access to networks and domains, it has very serious implications,” says Joseph Lorenzo Hall, a distinguished technologist at the Internet Society, a nonprofit that promotes an open internet. “It enables  fragmentation or what we call the ‘splinternet.’ Communications platforms are a light switch in the dark. Marginalized communities use TikTok in ways that aren’t just ‘ha ha, let’s do a dance.’”

In March, TikTok CEO Shou Zi Chew testified before the House Energy and Commerce Committee in a hearing that largely devolved into anxious, _Footloose_\-esque ruminations about child safety. Lawmakers raised concerns about TikTok that could be said of any social media platform and seemed to repeatedly, if unintentionally, make the case for comprehensive national privacy legislation in the US.

Meanwhile, even though Montana's ban will face legal challenges if passed, the very fact of its creation may embolden other states or even the federal government to consider paths to a TikTok ban, tipping off a chain reaction that could have profound impacts on freedom of digital expression in the US and around the world. 

“It’s a maddening irony that American legislators’ idea for countering China is to act more like China, home of the Great Firewall that censors its citizens’ free access to the flow of information,” Stanford's Pfefferkorn says. “Banning a popular social media app, especially on the basis of speculative concerns, is directly contrary to the vision of a free and open internet that the US has long promulgated abroad as part of our commitment to democracy.”

Montana’s Looming TikTok Ban Is a Dangerous Tipping Point

Security researchers are jailbreaking large language models to get around safety rules. Things could get much worse.

As a result, jailbreak authors have become more creative. The most prominent jailbreak was DAN, where ChatGPT was told to pretend it was a rogue AI model called Do Anything Now. This could, as the name implies, avoid OpenAI’s policies dictating that ChatGPT shouldn’t be used to produce illegal or harmful material. To date, people have created around a dozen different versions of DAN.

However, many of the latest jailbreaks involve combinations of methods—multiple characters, ever more complex backstories, translating text from one language to another, using elements of coding to generate outputs, and more. Albert says it has been harder to create jailbreaks for GPT-4 than the previous version of the model powering ChatGPT. However, some simple methods still exist, he claims. One recent technique Albert calls “text continuation” says a hero has been captured by a villain, and the prompt asks the text generator to continue explaining the villain’s plan.

When we tested the prompt, it failed to work, with ChatGPT saying it cannot engage in scenarios that promote violence. Meanwhile, the “universal” prompt created by Polyakov did work in ChatGPT. OpenAI, Google, and Microsoft did not directly respond to questions about the jailbreak created by Polyakov. Anthropic, which runs the Claude AI system, says the jailbreak “sometimes works” against Claude, and it is consistently improving its models.

“As we give these systems more and more power, and as they become more powerful themselves, it’s not just a novelty, that’s a security issue,” says Kai Greshake, a cybersecurity researcher who has been working on the security of LLMs. Greshake, along with other researchers, has demonstrated how LLMs can be impacted by text they are exposed to online through prompt injection attacks.

In one research paper published in February, reported on by Vice’s Motherboard, the researchers were able to show that an attacker can plant malicious instructions on a webpage; if Bing’s chat system is given access to the instructions, it follows them. The researchers used the technique in a controlled test to turn Bing Chat into a scammer that asked for people’s personal information. In a similar instance, Princeton’s Narayanan included invisible text on a website telling GPT-4 to include the word “cow” in a biography of him—it later did so when he tested the system.

“Now jailbreaks can happen not from the user,” says Sahar Abdelnabi, a researcher at the CISPA Helmholtz Center for Information Security in Germany, who worked on the research with Greshake. “Maybe another person will plan some jailbreaks, will plan some prompts that could be retrieved by the model and indirectly control how the models will behave.”

No Quick Fixes

Generative AI systems are on the edge of disrupting the economy and the way people work, from practicing law to creating a startup gold rush. However, those creating the technology are aware of the risks that jailbreaks and prompt injections could pose as more people gain access to these systems. Most companies use red-teaming, where a group of attackers tries to poke holes in a system before it is released. Generative AI development uses this approach, but it may not be enough.

Daniel Fabian, the red-team lead at Google, says the firm is “carefully addressing” jailbreaking and prompt injections on its LLMs—both offensively and defensively. Machine learning experts are included in its red-teaming, Fabian says, and the company’s vulnerability research grants cover jailbreaks and prompt injection attacks against Bard. “Techniques such as reinforcement learning from human feedback (RLHF), and fine-tuning on carefully curated datasets, are used to make our models more effective against attacks,” Fabian says.

Source

Wired