SectionAI Security
Attacks on and with AI: models, agents, data.
Anthropic: Claude Fable 5 Subscription Removal Is Temporary, Not Permanent
Anthropic and a Claude Code engineer have clarified that Fable 5 will return to standard subscription plans once capacity constraints ease, after…
Trump Administration Lifts Claude Restrictions After Safeguard Bypass Discovery
The Commerce Department has ended its ban on Anthropic's latest AI models following weeks of restrictions tied to a researcher-discovered method for…
Claude Fable Returns to Users, But Aggressive Safety Filters Draw Complaints
Anthropic has made Claude Fable 5 broadly available after a Commerce Department export ban was lifted, but users report the model is…
Microsoft Patches Vanishing Copilot Buttons in Classic Outlook
A service-side fix deployed June 29 restores missing Copilot UI elements for Classic Outlook users on the Copilot Chat (Basic) license. A…
Microsoft Teams Gets New Admin Controls to Block Unauthorized AI Bots
Microsoft has introduced a new Teams admin policy that forces explicit organizer approval before external bots can join meetings, addressing growing privacy…
AI Agent Runs Autonomous Ransomware Attack Against Production Database
Sysdig researchers say they have identified what may be the first ransomware attack executed end-to-end by an AI agent, with a large…
Six Questions to Pressure-Test Vendor Claims Around Frontier AI
As security vendors race to tout Frontier AI capabilities, enterprises need a sharper framework for separating genuine investment from marketing noise. A…
Fake Perplexity AI Extension Intercepted Search Queries via Chrome Web Store
A malicious Chrome extension impersonating Perplexity AI quietly rerouted all browser search queries through attacker-controlled infrastructure, Microsoft Threat Intelligence researchers found.
Poisoned MCP Tool Descriptions Can Turn AI Agents Into Data Exfiltration Tools
New Microsoft research demonstrates how attackers can manipulate AI agents into leaking sensitive company data by embedding malicious instructions inside tool descriptions,…
CIA Director Ratcliffe Outlines Aggressive Tech Overhaul at AWS Summit
CIA Director John Ratcliffe described sweeping organizational and procurement changes at the agency, framing frontier AI as a strategic imperative comparable in…
BioShocking Attack Tricks AI Browsers Into Bypassing Safety Guardrails
Researchers at LayerX demonstrated a prompt injection technique that uses a fictional game scenario to convince AI-powered browsers to exfiltrate sensitive data,…
Anthropic Launches Claude Sonnet 5 with Agentic Gains at a Lower Price
Anthropic's new Claude Sonnet 5 brings flagship-class agentic capabilities to a cheaper tier, narrowing the gap with Opus 4.8 while undercutting it…