Sign in Subscribe

AI Safety

Illustration of linked email and contact icons on a laptop, with Microsoft and Health-ISAC logos and an EvilTokens fraud headline.

Microsoft says EvilTokens turned stolen inboxes into AI fraud maps

Microsoft says EvilTokens used AI to turn compromised inboxes into fraud planning workflows, with more than 12,000 inboxes linked to the disrupted service.

Editorial illustration of a teen AI safety policy dashboard with Australia map cues and protective controls.

OpenAI brings teen AI safety into Australia's policy window

OpenAI published a six-pillar Australian Youth Safety Blueprint as the country consults on digital duty-of-care legislation.

Dark blue institutional briefing room illustration with an empty lectern, a glowing document folder and the words EU ASSESSES OPENAI REPORT.

OpenAI's agent incident has reached EU AI Act enforcement

The European Commission has received OpenAI's report on the German wiki incident, putting its disclosure response before EU AI Act officials.

Dark formal chamber with a blue-lit table, empty chair and unmarked folder, overlaid with the words Patients Should Know About AI.

UK AI-health report calls for patient notice and opt-outs

A UK Commission report asks healthcare organisations to give patients clearer notice of AI use and an opt-out where practical.

Dark school corridor with an unmarked classroom door and a blue-lit contract folder on a desk, overlaid with the words Schools Set Terms.

Schools get a contract playbook for AI data and oversight

A new school-AI contract standard gives districts a route to seek bans on training with student data, stronger oversight and clearer accountability.

Dark blue network of API request paths converging on a protected model gateway, with the words Track the Pattern.

U.S. cyber agencies want AI labs to spot distillation campaigns

NSA, FBI and CISA say model providers need coordinated detection for industrial-scale distillation campaigns that cross accounts, APIs and cloud services.

Dark cyber threat-field illustration with red arcs, a bright central node and the words Controls First.

Astra puts cyber controls in the release path

OpenAI says Astra has crossed its Critical cyber threshold, turning release access, monitoring and defensive-use limits into the main safety test.