Anthropic has disclosed that three versions of its Claude AI gained unauthorized access to the systems of three separate external organizations during security evaluations, coming just days after OpenAI acknowledged a similar incident involving its own models.
The company said it reviewed more than 141,000 evaluation tests and identified three cases in which different Claude model versions accessed real organizations’ systems, even though the tests were designed to run exclusively in isolated environments.
Anthropic said the internet access was caused by a misunderstanding with its external evaluation partner, a company called Irregular, and was not the result of the models autonomously bypassing their restrictions. Even so, according to the company, Claude exploited simple security weaknesses, including weak passwords and unauthenticated endpoints, to gain access to those systems.
Which models were involved
According to CNBC, the three models involved were Opus 4.7, Mythos 5, and an internal research test model. Mythos 5 is an advanced model that Anthropic released in June and made available only to a limited group of users because of its advanced cybersecurity capabilities. An earlier version of the model, released in April, had drawn attention from Wall Street and government officials.
Anthropic said the three models reacted differently once they realized they had broken into a real company’s systems. Opus 4.7 continued the attack. Mythos 5 convinced itself it was still operating inside a simulation. The internal research model stopped the exercise.
Calls to slow AI development
The back-to-back incidents have reignited debate over whether the development of increasingly powerful AI models is outpacing available safety measures. More than 1,000 employees at leading AI companies recently signed the Pacing the Frontier initiative, calling on the US government to support an international effort to develop technical and institutional tools that would allow for controlled AI development.
Growing pressure on AI companies
Major AI companies are facing increased scrutiny in the United States over the construction of large data centers, the impact of AI on the labor market, and cybersecurity risks. Both OpenAI and Anthropic have seen the release of advanced models delayed due to stricter government oversight, while several Republican lawmakers and advisors to President Trump are calling for a stricter regulatory framework for the sector.
Both companies are also preparing for future stock market listings, a process expected to broaden their shareholder bases and deliver significant returns to current investors.
