Anthropic's Claude Mythos outperformed GPT-5.5 in UK cybersecurity tests
Technology
Anthropic's new AI, Claude Mythos, has outperformed earlier results and OpenAI's GPT-5.5 in recent UK tests.
It managed to crack two tough cybersecurity challenges: it solved "The Last Ones" six out of 10 times and even tackled the previously unsolved "Cooling Tower" three times out of 10.
AISI: Mythos could complicate cyber defense
These results set a new standard for what AI can do in solving complex problems, but they also raise concerns: AISI points out that smarter AIs like Mythos could make defending against cyber threats trickier.
The tests were limited by token count, so with fewer restrictions, these AIs might be even more powerful than we realize.