AI models from Anthropic and OpenAI shake up Washington policy debate

New AI models from Anthropic and OpenAI are causing significant concern among Washington policymakers. Researchers who tested Anthropic's Mythos and OpenAI's GPT-5.5 describe their hacking capabilities as a 'game-changer'. The developments are prompting urgent discussions about AI regulation and national security.

Two cutting-edge artificial intelligence models developed by American tech giants Anthropic and OpenAI are sending shockwaves through Washington's policy circles, as researchers warn their capabilities represent a dramatic leap forward in what AI systems can do — including in the cybersecurity domain.

Anthropics's Mythos and OpenAI's GPT-5.5 have been put through rigorous testing by independent researchers, who concluded that the models' hacking capabilities are a "game-changer." The findings have alarmed lawmakers and security officials who are now scrambling to assess the national security implications of such powerful tools becoming widely available.

The revelations are intensifying an already heated debate in Washington about how to govern advanced AI systems. Policymakers on both sides of the aisle have long struggled to craft meaningful legislation that keeps pace with the breakneck speed of AI development, and the emergence of these new models adds fresh urgency to those efforts.

For Estonia and other NATO allies, developments of this nature carry direct relevance. As one of the world's most digitally advanced societies, Estonia has consistently been at the forefront of discussions around AI governance and cybersecurity policy at the European Union level. Estonian officials have previously stressed the importance of allied coordination on AI risk frameworks.