OpenAI Forms Mathematics Advisory Group as AI Models Crack Over 100 Unsolved Problems
OpenAI has established a formal advisory body focused on mathematical research, though the group will…
Tag: llm-security
BragJack Proof-of-Concept Shows One Malicious Extension Can Hijack AI Browser Agents Across Multiple Platforms
A researcher from Forever Security has demonstrated how a single rogue browser extension can subvert…
AI in Political Campaigning: Broad Listening Tools Offer a Different Model
While most campaigns use AI to broadcast messages at voters, a small number of civic…
Black Hat 2026: OpenAI and Hugging Face Reconstruct a Frontier Model Escape
A scheduled Black Hat USA 2026 session will walk through how a frontier AI model…
Anthropic Documents AI-Assisted Weapons Development by Threat Actors in Yemen
Claude was used as a software engineering team substitute to develop guidance, navigation, and control…
When AI Becomes the Attacker: Schneier’s DEF CON Address on Autonomous Hacking
Security veteran Bruce Schneier used DEF CON to lay out what the security industry needs…
San Francisco Demands Meta Explain How Child Exploitation Ads Ran on Facebook and Instagram
A city attorney's order puts Meta's ad moderation practices under formal scrutiny, raising platform governance…
NCSC Flags Shadow AI as an Emerging Corporate Data Risk
The UK's National Cyber Security Centre has warned that employees using unapproved AI tools are…
OpenAI’s Internal Shift Toward Agentic Engineering Raises Governance Questions for Security Leaders
OpenAI researchers are now routinely using coding agents in their own work, signalling a broader…
Autonomous AI Agents Are Probing Your Perimeter and Logging What They Find
Field reports from self-directed AI agents reveal that identity verification is rarely the barrier it…