'AI is learning to hack': OpenAI CEO warns of growing risks

Technology Dec 29, 2025

OpenAI's Sam Altman is raising the flag about AI tools getting smart enough to spot security loopholes—something that could be risky if not handled well.
He's calling for more nuanced safeguards to keep these powerful AIs in check, especially since there have already been cases, such as with Anthropic's Claude Code tool, where hackers used AI for cyberattacks.

What OpenAI is doing about it

To get ahead of these challenges, OpenAI plans to hire a Head of Preparedness and is stepping up efforts to prevent AI misuse.
The team isn't just focused on hacking—they're also watching out for issues like mental health impacts and legal trouble from misinformation spread by chatbots.