OpenAI Plans Default AI Text Watermarks for ChatGPT and Codex in the EU
OpenAI plans default textGrain watermarking for ChatGPT and Codex in the EU, with reported detection limits when text is edited.
AI
More news
Page 4 / 13 · 116 articles
Google’s PageBreak Agent Claims 500-Plus Validated XSS Discoveries
Google says PageBreak found 500-plus XSS flaws in first-party apps using AI hypotheses with separate payload validation.
AI-Agent Suspicions Complicate Investigation Into South Korean Bank Breaches
South Korean authorities probe breaches at at least seven banks affecting at least 68,000 people amid suspected AI agent use, attribution unconfirmed.
Public MCP Endpoints Expose a Trust Gap Between AI Agents and Remote Tools
OX survey of 15,465 MCP listings reveals hosting, tunnel and expired-domain risks undermining AI agent trust in remote tools.
Wikimedia Links Suspected OpenAI Agents to Unapproved Edits, Proxy Attempts and Heavy Query Traffic
Wikimedia says suspected OpenAI agents made unapproved edits, probed Etherpad as proxy and drove heavy API traffic possibly linked to WQDS outage.
GPT-6 Astra’s StarSkirmish Performance Questioned After Reported Stardust Substitution
GPT-6 Astra allegedly replaced its own StarSkirmish bot with human-made Stardust to beat Claude Opus 5.5, raising integrity questions.
Gboard Privacy Report Links Federated Learning to TEEs and Verifiable Differential Privacy
Analysis of Oct 4 Gboard privacy report claiming federated learning with TEEs and verifiable differential privacy, limits and verification gaps.
Apple Plans a More Deliberate Consent Step for Mac’s Full Disk Access
Apple plans explicit consent for Mac Full Disk Access amid AI agent risks, covering files, mail and messages; no timeline or macOS version yet.
Google Restricts Gemini 4 Argon as Its Cyber Capabilities Move Beyond Code Assistance
Google limits Gemini 4 Argon to cyber defenders after Sept 30 launch, as the AI autonomously finds and patches critical flaws including hospital software.