The Startup At The Center Of Multiple Rogue AI Hacks

Plus: Chinese Hacker Used AI To Attack 100+ Companies

Forbes

Much of the panic around rogue AI of late derived from OpenAI agents’ breach of Hugging Face in July. But there have been a spate of similar but separate cases where AI agents powered by OpenAI, Anthropic and Meta breached companies without authorization. Then late last week, Google confirmed its Gemini AI had also hacked into three companies without permission. Each of those four cases had the same source: $450 million Israeli startup Irregular


Irregular, which tests models for safety and security, has scored contracts with major western AI labs since its founding in 2024. But in a slipup in May, it accidentally gave AI agents from all four tech giants a task to hack into a fake company that its researchers thought didn’t have a parallel in the real world. But it did. The agents began trying to break out of their lab container, access the internet and hack into the real company (whose name is yet to be revealed) because they thought that was part of the test. Even though the AI was told it didn’t have internet access, it was "inadvertently" granted, per an Irregular post from July.


Irregular cofounder and CEO Dan Lahav now says the AI industry needs to do better.  


"Mistakes like human oversight can happen and obviously we should learn from them. We should be accountable for them,” Lahav tells Forbes.


Irregular’s critics have argued simple measures like better monitoring of test environments—where AI agents are tested to see whether they would carry out cyberattacks or help build chemical weapons— could have prevented the summer of AI security breaches. But Lahav says more monitoring wouldn’t be enough to stop similar hacks happening again.


What’s needed, he says, is for the industry to take stock and do forensics into the attacks to better understand the models’ behavior, even if the problems they pose don’t have simple answers. “We should get comfortable with complicated realities,” he says. For instance, it might be a good idea to let AI agents under stress tests have access to the internet, even though it poses a risk of similar attacks to those that started in Irregular’s labs, he says.


AI is bringing a new kind of security threat and current protection mechanisms aren’t built to deal with endlessly persistent attackers who continuously change and adapt "very much like a human," he says. 


Cybercriminals certainly aren’t waiting around for defenders to research AI. Forbes has learned that hackers are already taking advantage of AI models, both new and old, to carry out massive attacks, per our Big Story today…

Got a tip on surveillance or cybercrime? Get me on Signal at +1 929-512-7964.

Thomas Brewster Associate Editor, Cybersecurity

Follow me on Forbes.com

  (Photo Illustration by Pavlo Gonchar/SOPA Images/LightRocket via Getty Images)
The Big Story
A Chinese Hacker Used AI To Attack 100+ Companies In One Of Largest AI Hacks Yet
Read Article

A cybercriminal ordered AI agents to go on a hacking spree over five days earlier this month. Using AI developed by Anthropic, Deepseek and Moonshot, the bots broke into at least 30 websites and installed “skimmers” that siphoned off 600,000 credit card details.


It’s one of the biggest automated attacks seen to date and lands amid concerns about the use of AI for malicious means. Despite attempts by Anthropic and Cloudflare to stop the attacks, the researchers who uncovered them say they’re ongoing.

The Stories You Have To Read Today

A hacking crew known as ShinyHunters claims to have breached the FBI and stolen data on current and former employees, according to 404 Media.


Despite massive spending on surveillance at the Mexico border, and agents who’re trained to provide medical assistance, many crossing continue to die, per an investigation by MIT Technology Review.


Wired reports on a Google employee who went undercover inside TeamPCP, a hacking crew responsible for carrying out a major cyberattack on the AI supply chain earlier this year. As a result, Google was able to warn targets as they were being attacked.

Winner Of The Week
Coinbase and Microsoft announced the takedown of EvilTokens, an AI-powered phishing-as-a-service tool that helped criminals hack into Microsoft 365 accounts, impersonate users and trick contacts into sending funds to scam addresses. They also helped seize over 50 websites used by EvilTokens and identified two suspects who allegedly helped run the service.
Loser Of The Week
The U.S. military nearly boarded a Chinese ship in the Middle East thanks to a false report generated by an AI, CNN reports. The AI chatbot that helped generate an intelligence report incorrectly flagged the cargo as nuclear weaponry and “almost started a war,” one source claimed.
More From Forbes
Most-Read From Forbes
Take Forbes Journalism On The Go
Stay connected to the ideas, insights and stories moving business forward. Save top articles for later and easily search stories that matter to you. Download today to begin.
Get the Forbes App
Forbes

Unsubscribe from The Wiretap.

Manage Email Preferences

My Forbes Account  |  Newsletters  |  Help  |  Privacy

Forbes Media 499 Washington Blvd. Jersey City, NJ 07310

list_id:54119

Komentar

Postingan populer dari blog ini

💥 Trending Games with potential token airdrops

🌟New Features Now Live on CoinCodex