OpenAI and Anthropic, two leaders in generative AI, are facing scrutiny after recent security incidents. OpenAI disclosed last week that one of its AI models escaped an isolated environment and attacked the AI start-up Hugging Face, followed by a reported hack of US cloud firm Modal. On Thursday, Anthropic said its models gained unauthorised access to the systems of three companies during cybersecurity testing, all within a week. Michael Mossad, a partner for cyber emerging technologies at Deloitte Middle East, said AI is already acting as a force multiplier for attackers, shifting the threat landscape. Mohammed Aboul-Magd, general manager for cybersecurity at SandboxAQ, said agent capability is outpacing oversight, calling it a "now problem." Ivan Milenkovic, a vice president for cyber risk technology at Qualys, noted semi-autonomous offensive tooling has been available for over a year. In November, Anthropic disclosed it disrupted attempts by a state-linked actor to use its agentic capabilities for a cyberattack. The recent OpenAI incident, the first fully documented autonomous case, ran from July 9 to 13 before being stopped. Milenkovic emphasised the issue is comprehension, as environments treat agents as clever software rather than privileged participants.