Live Roundups Editor's Desk Insights Executive Ops Weather About 🔍
Home › Technology › OpenAI and Anthropic models ‘went rogue’ during UK...
Technology

OpenAI and Anthropic models ‘went rogue’ during UK cybersecurity test

The Guardian Tech Wednesday 05 August 2026, 08:40 UTC By Dan Milmo Global technology editor 1 min read
Key Points

AI Security Institute says tools engaged in potentially harmful activity and incident reveals new type of riskAdvanced AI models developed by OpenAI and Anthropic went rogue during a cybersecurity test and showed a new type of risk posed by the technology, according to the UK’s AI Security Institute. AISI described the actions carried out by the agents, the term for AI systems that can perform tasks without human help, as a “serious incident”. In one example, an agent powered by Anthropic’s...

AI Security Institute says tools engaged in potentially harmful activity and incident reveals new type of risk

Advanced AI models developed by OpenAI and Anthropic went rogue during a cybersecurity test and showed a new type of risk posed by the technology, according to the UK’s AI Security Institute.

AISI described the actions carried out by the agents, the term for AI systems that can perform tasks without human help, as a “serious incident”. In one example, an agent powered by Anthropic’s Mythos model sent targeted emails to people.

Continue reading...
UK (LOCATION) AI Security Institute (ORG) AI (ORG) AISI (PERSON)
Originally published by The Guardian Tech Read original →

Related Stories

German airport diverts flights after security alert over unidentified object

German airport diverts flights after security alert over unidentified object Bomb disposal robot deployed at Leipzig-Halle as investigators probe potential threat to cargo hub German police said Wednesday that a “security related” incident involving an unidentified object found at an airport forced the diversion of flights and the deployment of a bomb disposal robot overnight. The Bild newspaper said the object found at Leipzig-Halle airport was a drone with a detonator and possibly...

South China Morning Post 4m ago

OpenAI and Anthropic models went on a hacking spree when tested by the UK's AI research institute

OpenAI and Anthropic models went on a hacking spree when tested by the UK's AI research institute The AI agents even used social engineering techniques and left instructions for future agents. Both OpenAI and Anthropic recently admitted that their models escaped from their testing environments and hacked into outside organizations on their own. Now, the UK's AI Security Institute (AISI) has released a report, detailing how the companies' models also acted independently and "engaged in...

Engadget 21m ago

Sandisk Earnings Give Dip Buyers an Opening After Stock’s Plunge

Sandisk Earnings Give Dip Buyers an Opening After Stock’s Plunge Sandisk Corp.’s high flying stock tumbled back to Earth last month. Its earnings after the market closes Wednesday will test whether strong results are enough to entice investors to buy in again. Shares of the maker of memory storage components plunged 47% in July, erasing more than $150 billion in market value in their worst monthly performance since going public in February 2025.

Bloomberg Technology 29m ago

Germany news: Lightning strike cuts Bavarian town in two

Lightning strike cuts Bavarian town in two Published August 5, 2026last updated August 5, 2026What you need to know - A lightning strike disabled two railway crossing barriers, effectively splitting a German town in two - A German employers' organization warns against the watering down of government pension reforms - Flights at Leipzig/Halle Airport were briefly suspended after an unidentified object was spotted near a runway Here is a roundup of top stories from and about Germany on...

Deutsche Welle 33m ago