OpenAI pledges to add Astra security as Anthropic loosens Fable's leash

REG AD AI and ML Or how I learned to stop worrying and love dangerous AI After acknowledging last month that unreleased AI models committed what for human perpetrators would be computer crimes, OpenAI now says it cannot rule out the possibility that Astra, a pending model release not involved in its Hugging Face hack,

OpenAI said it has suspended work on some aspects of its upcoming model Astra over concerns about its cybersecurity prowess.

Two advanced AI systems from OpenAI and Anthropic created fraudulent human profiles and attempted to deceive people in simulated cyberattacks during testing conducted by the UK's AI Security Institute.

‘Some of the agents being tested had engaged in sustained, potentially harmful activity directed at real people and organisations,’ says UK’s AI safety watchdog

The institute said agents powered by Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol engaged in unauthorized actions during security evaluations the government organization conducted.

'Potentially harmful activity': OpenAI, Anthropic AI agents implicated in new security breaches  dawn.com

Anthropic and OpenAI's agents engaged in unauthorised actions during security evaluations the government organization conducted to assess the models' capabilities.

SAN FRANCISCO, Aug 4 - An AI agent was caught creating fake online identities to gain unauthorized access to secure systems during tests of models from OpenAI and Anthropic which revealed a series of new breaches, Britain's AI Security Institute (AISI) disclosed on Tuesday.

Both companies acknowledge the incidents and express commitment to improving safety practices in AI evaluations

Report underscores lax state of safeguards around agents being marketing as future of business

SAN FRANCISCO, Aug 4 : An AI agent was caught creating fake online identities to gain unauthorized access to secure systems during tests of models from OpenAI and Anthropic which revealed a series of new breaches, Britain's AI Security Institute (AISI) disclosed on Tuesday.The institute said agents powered by Anthropic

OpenAI, Anthropic AI agents implicated in new security breaches  Reuters

Meta, Anthropic, Google and OpenAI staff will meet with U.S. President Donald Trump’s advisers on Tuesday about voluntary safety testing for advanced AI models, according to four sources familiar with the meeting, as concerns over rogue AI agents grow.

Hackers are using advanced generative AI models to create malware and automate cyberattacks. Researchers found threat actors easily bypass AI safety guardrails using simple social engineering tactics. These same AI capabilities designed for security are now being manipulated by malicious actors. Attackers also leverage

Anthropic and OpenAI disclosed in recent days that their AI tools breached the systems ‌of ‌other companies, stirring concerns among U.S. lawmakers.

If you were to strip out Big Tech's investment gains from private AI companies, the earnings story is a lot less bullish.