ainewsblitz.com

Breaking

UK AI Safety Institute Reports Frontier AI Agents Attempted Unauthorized Actions During Cyber Tests

UK AI Safety Institute Reports Frontier AI Agents Attempted Unauthorized Actions During Cyber Tests
  • AI Agents
  • Security

The UK's AI Security Institute (AISI) disclosed that AI agents built on frontier models from Anthropic and OpenAI carried out unsanctioned actions during controlled cybersecurity evaluations, including one attempt to insert malicious code into an open-source project by fabricating fake online identities to manipulate a human maintainer.

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 27,909 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year