Technology

OpenAI Overhauls Safety Protocols After Its AI Agents Went Rogue

Wired UK · 18 August 2026 at 19:33

OpenAI Overhauls Safety Protocols After Its AI Agents Went Rogue

The ChatGPT maker says its upcoming Astra model may have reached “critical” cyber capabilities, prompting it to halt a significant number of training runs while it tightens internal safeguards.

Read the full article on Wired UK ›

5News aggregated this summary from the outlet’s public feed. The full article, with all the context, is on www.wired.co.uk — the content belongs to Wired UK.

OpenAI Overhauls Safety Protocols After Its AI Agents Went Rogue

More from Wired UK

Garmin Watches Are Up to $250 Off Right Now On Amazon (2026)

Viral Disneyland Content Creator Defends Her ‘Special Friendship’ With Peter Pan

Meta Ran Ads for an App That Promised to Nudify Female Politicians

Can the Upcoming ‘Expanse’ Game Avoid the Biggest Mistake of ‘Mass Effect’?

More in Technology

North Face everyday backpacks are up to 33% off at Amazon — just in time for back to school season

'This technology turns every router into a potential means for surveillance': Report claims Wi-Fi devices could 'quietly identify' people with nearly 100% accuracy

The iPad Air M4 hits its biggest discount yet since Apple's recent price hikes — save $100 at Amazon

Garmin Watches Are Up to $250 Off Right Now On Amazon (2026)

Cherokee Nation joins list of tribes banning data centers on tribal lands due to water, energy, noise, and cultural resource protection concerns — and all new projects require ‘early consultation’

I tried ChatGPT's new interactive quizzes on 5 subjects I thought I knew well — it quickly found the gaps in my knowledge