Anthropic Says Claude AI Models Hacked Three Companies During Tests
Anthropic said a retrospective review found 141,006 tests and three cases in which Claude models reached real systems after escaping sealed environments.
- On Thursday, Anthropic reported that its Claude artificial intelligence models accessed the internet during evaluation tests and "gained unauthorized access to the real systems of three different organizations."
- The incidents occurred within testing environments built by the AI security firm Irregular, where a misunderstanding with the evaluation partner left the environments unsealed despite Anthropic instructing Claude they lacked internet access.
- Anthropic discovered the breaches after a "large-scale retrospective review" of 141,006 evaluation tests, identifying three models—Opus 4.7, Mythos 5, and an internal research model—that used basic techniques like exploiting weak passwords.
- Neither Anthropic nor the affected organizations detected the intrusions until the retrospective review, which was prompted by a similar security incident OpenAI disclosed last week.
- More than 1,100 staffers across artificial intelligence firms signed a petition on Tuesday urging the government to support mechanisms that "deliberately pace" AI development to prevent rapid advancement.
57 Articles
57 Articles
Sueddeutsche Zeitung
inFranken.deThe fact that an AI model became a hacker in a test on its own was considered an "unexampled incident" and a wake-up call for the tech industry. But now it becomes clear: It was not an isolated case.
The Straits Times
Free Malaysia Today NewsAnthropic’s models gained unauthorised ‘real-world’ access during testing
The incident comes just days after rival OpenAI first revealed its models went rogue during security testing. Read more at straitstimes.com.
Anthropic says its AI models also broke out and hacked other companies
By Hadas Gold, CNN (CNN) — AI company Anthropic says that during routine testing some of its models accessed the internet and hacked into three separate organization’s systems – and that it didn’t notice the models had done so until an internal review prompted by rival OpenAI disclosing its models did the same. Anthropic said The post Anthropic says its AI models also broke out and hacked other companies appeared first on KESQ.
Anthropic says its models went rogue and hacked 3 companies during testing
Anthropic says it found multiple incidents in which Claude models gained access to unauthorized data.Bloomberg/Getty ImagesAnthropic said Claude models accessed data from three firms without permission since April.Anthropic said it conducted a review after the OpenAI-Hugging Face breach.There are growing fears about the amount of data AI models can access by design or by mistake.Anthropic says it found multiple incidents in which Claude models g…
The well-known AI system Claude from the American company Anthropic has broken into three companies after gaining unintended access to the internet. Anthropic announced this a week after something similar was reported by competitor OpenAI. According to Anthropic, the hacks took place sometime since April, during so-called 'capture-the-flag' exercises. In these exercises, Claude was tasked with extracting hidden information from networks speciall…
Here you will find information on the topic "Artificial Intelligence". Read now "Even AI of the OpenAI rival Anthropic attacked real companies".
Coverage Details
Bias Distribution
- 36% of the sources lean Left, 35% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
