AI just built its first zero-day exploit

The New York Times

OpenAI is weighing legal action against Apple over the disappointing integration of ChatGPT into Siri and Apple devices, straining a partnership that began two years ago. The potential lawsuit would add to OpenAI's legal challenges as it simultaneously defends against Elon Musk's multibillion-dollar claim.

Bloomberg

Google detected the first instance of criminals using AI to autonomously discover and weaponize a zero-day exploit targeting 2FA in web admin tools, marking a major escalation in AI-powered cyber threats. The detection disrupted a planned mass exploitation campaign.

CNBC

The U.S. and China established a formal AI safety protocol at the Trump-Xi Beijing summit, with Treasury Secretary Bessent asserting America's negotiating leverage stems from its leadership in AI development. The agreement focuses on best practices and preventing nonstate actors from obtaining advanced AI models amid ongoing semiconductor export tensions.

Anthropic

Anthropic is partnering with the Gates Foundation in a $200 million commitment to apply Claude AI to global health, education, and economic mobility challenges, focusing on vaccine development for neglected diseases and AI-powered educational tools in underserved regions. The initiative aims to extend AI benefits beyond commercial markets through grants, technical support, and open-access public goods.

9to5Mac

Security researchers leveraged Anthropic's Mythos Preview to bypass Apple's five-year, billion-dollar Memory Integrity Enforcement system on M5 Macs in just five days, demonstrating how AI models can dramatically accelerate vulnerability discovery and exploit development even against state-of-the-art hardware security mitigations.