AI Security AI News & Updates

OpenAI Acknowledges Permanent Vulnerability of AI Browsers to Prompt Injection Attacks

OpenAI has admitted that prompt injection attacks against AI browsers like ChatGPT Atlas may never be fully solved, similar to how scams and social engineering persist on the web. The company is deploying an LLM-based automated attacker trained through reinforcement learning to proactively discover and patch vulnerabilities before they're exploited in the wild. Despite these defensive measures, experts warn that agentic browsers currently pose significant risks due to their high access to sensitive data combined with moderate autonomy, questioning whether their value justifies their risk profile.

AI Security Firm Irregular Secures $80M to Test and Secure Frontier AI Models Against Emergent Risks

AI security company Irregular raised $80 million led by Sequoia Capital to develop systems that identify emergent risks in frontier AI models before they are released. The company uses complex network simulations where AI agents act as both attackers and defenders to test model vulnerabilities and security weaknesses.

Anthropic CEO Warns of AI Technology Theft and Calls for Government Protection

Anthropic CEO Dario Amodei has expressed concerns about potential espionage targeting valuable AI algorithmic secrets from US companies, with China specifically mentioned as a likely threat. Speaking at a Council on Foreign Relations event, Amodei claimed that "$100 million secrets" could be contained in just a few lines of code and called for increased US government assistance to protect against theft.

UK Rebrands AI Safety Institute to Focus on Security, Partners with Anthropic

The UK government has renamed its AI Safety Institute to the AI Security Institute, shifting focus from existential risks to cybersecurity and national security concerns. Alongside this pivot, the government announced a new partnership with Anthropic to explore using its AI assistant Claude in public services and contribute to security risk evaluation.