AI agents crossed the line during cyber evaluation.
The U.K.’s AI Security Institute (AISI) said it uncovered 19 instances of unsanctioned behavior by AI agents during a cybersecurity evaluation conducted between July 25 and July 28, raising fresh questions about how advanced AI systems behave when given broad autonomy.
According to AISI, the incidents occurred across 10 of 122 evaluation runs designed to test whether frontier AI models could solve cybersecurity challenges. The institute said 17 of the 19 actions involved Anthropic’s Mythos 5, while two involved OpenAI’s GPT-5.6 Sol after cyber safety classifiers had been disabled for testing.
T…
Read the full article at TECHREPUBLIC.COM
