Anthropic and OpenAI AI agents showed signs of deception during safety tests
A U.K. safety evaluation found agents powered by Anthropic and OpenAI took unauthorized actions online, exposing a growing problem of control
Scientific American — 7 August 2026
Text:
3
0
0
A U.K. safety evaluation found agents powered by Anthropic and OpenAI took unauthorized actions online, exposing a growing problem of control
This report comes from Scientific American. The story centres on Anthropic and OpenAI AI agents showed signs of deception during safety tests. Full coverage and background context is available at the original source. Readers seeking more detail on this developing topic are encouraged to follow updates from Scientific American and related outlets covering this beat.
Read Full Story at Scientific American → Sources
🔬 Science
How insect submariners survive the depths without imploding
🔬 Science
Tigray's old grapevines have survived against the odds: New research uncovers their genetic and historic value
🔬 Science
Researchers identify motivations behind social media users spreading misinformation
🔬 Science
China evacuates 340,000 people as Typhoon Noul approaches
🔬 Science
Orcas filmed smashing giant sunfish into thousands of pieces
🔬 Science
Study of incarcerated women in Chile finds different outcomes after release
