Wed, Aug 5 03:23 AM

AI used new levels of 'autonomy and deception' to trick people in safety test

BBC News•Wed, Aug 5, 2026, 12:02 AM•2 min read

The UK's AI Safety Institute said recent behaviour from Anthropic and OpenAI models was malicious and unprecedented.

Source: [BBC News](https://www.bbc.co.uk/news/articles/c1w1lvn7d9go?at_medium=RSS&at_campaign=rss)

📰 Read Full Story

This is an aggregated headline summary. For the complete report, visit the original publisher.

Continue Reading at BBC News ↗

#ai #safety #used #levels #autonomy #deception #trick #people #test

More Headlines

FinanceInvesting.com• 34m ago

Exclusive-Samsung, SK Hynix test Chinese chip tools as hedge against US risks

Developing story — details emerging. Check the source link for the latest updates.

AI & MLInvesting.com• 59m ago

OpenAI, Anthropic AI agents implicated in new security breaches

Developing story — details emerging. Check the source link for the latest updates.

TechnologyHacker News• 1h ago

AI used new levels of 'autonomy and deception'

4 points, 0 comments on Hacker News

TechnologyHacker News• 2h ago

To Save Japan's Heritage Homes, People Are Shipping Them Across the World

2 points, 0 comments on Hacker News

AI & MLNew York Times• 3h ago

Trump White House Readies AI Framework to Review Security Risks

The voluntary review process will cover closed-source artificial intelligence models, but exclude those that publish the underlying code.

SciencePhys.org• 3h ago

Why life's hardships may make conspiracy theories more appealing

People who experience greater hardship throughout their lives may become more susceptible to conspiracy theories when those experiences lead them to see the world as fundamentally unfair, according to new research.