Anthropic Expands Project Glasswing to 150 Partners Across 15 Countries
Anthropic has scaled Project Glasswing to 150 partners in 15 countries, finding over 10,000 critical software vulnerabilities.
Safety
176 articles
AI safety, alignment, and governance — model risk, red-teaming, regulation, and policy. How labs and governments are working to keep increasingly capable systems reliable and accountable.
LATEST NEWS
-
-
OpenAI Proposes Global Youth AI Safety Institute
OpenAI calls for an international youth safety institute to ensure safe and age-appropriate AI access for young people, as G7 leaders prepare to discuss the issue in June 2026.
-
ZeroDrift Raises $10 Million to Protect AI Models From Themselves
ZeroDrift, an AI compliance service, raised $10 million in seed funding to safeguard AI models from generating non-compliant content.
-
IEEE Calls for Measuring AI's Societal Impact Beyond Performance
IEEE experts argue that current AI metrics focus too much on performance, not on how these models affect human society.
-
White House Struggles to Revive AI Regulation Plan
The White House is facing internal conflict over reviving an AI regulation plan canceled by Trump in May, with uncertainty about its future.
-
Meta Unveils Advanced AI Scaling Framework for Safety
Meta released its updated Advanced AI Scaling Framework and Safety & Preparedness Report for Muse Spark on April 8, 2026, detailing new safety measures for advanced models.
-
Meta AI Support Chatbot Vulnerable to Hacker Exploits
Hackers exploited Meta's AI chatbot to take over notable Instagram accounts worth hundreds of thousands of dollars before a May 29 patch.
-
Florida Sues OpenAI and Sam Altman Over ChatGPT's Links to Violent Incidents
Florida filed a lawsuit against OpenAI and its CEO, Sam Altman, alleging their negligence contributed to violent incidents involving ChatGPT. The case is the first state-led legal action against the company over such claims.
-
Amazon Introduces Policy and Lambda Interceptors for Secure AI Agents in Bedrock AgentCore
Amazon Bedrock AgentCore now supports policy-based access control and Lambda interceptors to secure AI agents, enabling dynamic validation and governance for enterprise workflows.
-
Author of AI Truth Book Uses AI Tools in Writing
WIRED's excerpt from Steve Rosenbaum's book, which explores AI's impact on truth, was flagged as possibly AI-generated, raising questions about its authenticity.
-
Anthropic's Olah Speaks at Vatican on AI Ethics
Chris Olah, cofounder of Anthropic, addressed the Vatican on AI ethics following Pope Leo's encyclical, emphasizing the need for industry restraint.
-
Illinois Passes Major AI Safety Law for Frontier Labs
Illinois House passes SB 315, requiring third-party audits of AI safety practices by major labs like OpenAI and Google DeepMind.