An AI startup has claimed that it hacked Anthropic’s systems in less than 12 hours, exposing internal messages and data.
US-based Cathedral, which was founded this year by former staff members of the US Department of Government Efficiency, said it breached the artificial intelligence company’s internal network as part of a bug bounty program that encourages outsiders to look for security flaws.
First reported by the Washington Post, the breach exposed Slack messages, meeting transcripts, staff salaries and job offers.
Anthropic confirmed that Cathedral had reached a “limited set” of sales, HR and IT systems, but did not access users’ conversations with Claude. The AI model’s source code was also not compromised.
“Our monitoring flagged some of Cathedral’s activity before they reported it to us,” Anthropic said in a statement shared with the publication. “We shut off their access the same day.”
A senior Trump administration official, speaking anonymously, said: “If a bunch of kids who have just started a company can do this, then potentially the Chinese aren’t far behind.”
Anthropic pushed back on the claim that it only took 12 hours to breach the systems, with a spokesperson telling The Independent that it “involved attempts over several days”.
The spokesperson added that Cathedral’s actions went “substantially beyond” the company’s bug bounty program by targeting third party applications.
Security researchers noted that the incident was not a sophisticated AI attack, but a basic “hygiene failure” with third party software used by Anthropic.
“This probe by Cathedral shows how even one of the most innovative companies can still be susceptible to a supply-chain attack,” Heath Mullins, the Chief Evangelist at cyber security firm ExtraHop, told The Independent.
“AI providers are becoming a primary target because so many organizations now build their workflows on top of them. When a handful of platforms sit at the center of how a majority of businesses operate, a single compromise becomes a supply chain problem for everyone downstream.”
The revelation came just hours after Anthropic filed for its initial public offering (IPO), which laid out the firm’s ambition to build AI systems capable of curing cancer and reshaping the global economy.
Anthropic said in the filing that its AI is “one of the most consequential technologies in human history”, capable of “potentially curing diseases that have plagued humanity for millennia, including cancer Alzheimer’s disease, heart disease, rare genetic diseases and major infectious diseases.”
The filing also noted that Anthropic’s AI models may pose “catastrophic or existential risks to humanity” due to how powerful they are.
The IPO is expected to be the biggest listing in history, exceeding the $1.8 trillion valuation of Elon Musk’s SpaceX.
Read MoreChina unveils 13-foot robot that can lift a cow
California city weighs data center ban over fears they could send power bills soaring
Anthropic has opened its books to the stock market - they make for terrifying reading
‘Skynet’ AI hijacks other AI to carry out attacks, researchers warn
OpenAI starts training of powerful systems that experts fear could go rogue
What you need to do as the world panics about AI cyber attacks
