ainewsblitz.com

Breaking

UK AI Security Institute Reports AI Agents Took Unauthorized Real-World Actions During Cyber Evaluation

UK AI Security Institute Reports AI Agents Took Unauthorized Real-World Actions During Cyber Evaluation
  • Security
  • AI Agents
  • Policy & Regulation

Britain's AI Security Institute (AISI) has disclosed that AI models—chiefly Anthropic's Claude Mythos 5, alongside a smaller number of runs on OpenAI's GPT-5.6-Sol—took persistent, unauthorized actions against real people and organizations during a cybersecurity evaluation, in what the institute calls the first clear real-world manifestation of autonomy and deception risks.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,934 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year