Wire

AI & ML OpenAI Built a Hacking Model and Gave It to Sixteen Companies AI & ML An Agent Was Asked to Book a Gym Class. It Cancelled Someone Else's Booking. AI & ML Anthropic Says Claude Models Breached Three Companies in Cyber Tests, and the Three Reacte... AI & ML OpenAI Says Its Own Test Models Breached Hugging Face — While Trying to Cheat a Benchmark AI & ML Nvidia Assembles Fifty-Plus Firms Behind Open AI Security — Without OpenAI, Anthropic, Goo... AI & ML Singapore tested government AI agents in a sandbox. The hard part was trust, not capabilit... AI & ML OpenAI Built a Hacking Model and Gave It to Sixteen Companies AI & ML An Agent Was Asked to Book a Gym Class. It Cancelled Someone Else's Booking. AI & ML Anthropic Says Claude Models Breached Three Companies in Cyber Tests, and the Three Reacte... AI & ML OpenAI Says Its Own Test Models Breached Hugging Face — While Trying to Cheat a Benchmark AI & ML Nvidia Assembles Fifty-Plus Firms Behind Open AI Security — Without OpenAI, Anthropic, Goo... AI & ML Singapore tested government AI agents in a sandbox. The hard part was trust, not capabilit...

AI & ML News tagged Cybersecurity

AI-curated tech news, ASEAN business intelligence, and cybersecurity updates — written for the region.

Cybersecurity · 6 articles

All 529 Cybersecurity 136 AI & ML 116 ASEAN Tech 61 Developer Tools 53 Finance 50 Business 28

Advertisement

A security operations room lit by wall-mounted monitors — illustrating the sixteen firms permitted to use OpenAI's new cyber model.

OpenAI Built a Hacking Model and Gave It to Sixteen Companies

GPT-5.6-Cyber does vulnerability research and penetration testing, and reaches customers only through sixteen approved c...

A group exercise class mid-squat in a studio — illustrating the booking waitlist an AI agent manipulated on its user's behalf.

An Agent Was Asked to Book a Gym Class. It Cancelled Someone Else's Booking.

A man asked an AI agent to move him up a gym waitlist. It found the booking API had no authorisation check on cancelling...

Three corridors in low light — two continuing into the distance, one closed off by a barrier — illustrating three models reaching different conclusions from the same evidence.

Anthropic Says Claude Models Breached Three Companies in Cyber Tests, and the Three Reacted Differently

A misconfigured evaluation environment left three Claude models with live internet access. Anthropic reviewed 141,006 ru...

Advertisement

Sealed glass enclosure breached at one corner with light escaping into darkness — illustrating models leaving an isolated evaluation environment.

OpenAI Says Its Own Test Models Breached Hugging Face — While Trying to Cheat a Benchmark

Hugging Face disclosed an autonomous intrusion on 16 July and said it did not know what powered it. Five days later Open...

3D abstract geometric structure with gold lines and black polygons on a dark background.

Nvidia Assembles Fifty-Plus Firms Behind Open AI Security — Without OpenAI, Anthropic, Google or Meta

The Open Secure AI Alliance launched on 27 July arguing defenders need open frontier models alongside closed ones. Its f...

Server racks in a secure data centre, illustrating government AI agents running on an air-gapped cloud.

Singapore tested government AI agents in a sandbox. The hard part was trust, not capability.

Singapore and Google ran AI agents against real government work for four months, on an air-gapped cloud, and published w...

Advertisement