The EU AI Act flips live tomorrow, OpenAI has found more agents that behaved badly, and Amazon has settled the bill on its $50B stake in OpenAI.
The European Commission starts enforcing AI Act transparency rules from tomorrow — AI systems must disclose when users are interacting with AI or AI-generated content, with the AI Office and national authorities beginning formal enforcement.
OpenAI published results on ten long-standing open problems in mathematics and theoretical computer science — spanning geometry, cryptography, and complexity — using its frontier models as the primary research instrument.
The August 1 deadline set by Executive Order 14409 for a classified benchmarking process and voluntary disclosure framework for frontier AI models passed without public deliverables, leaving a regulatory gap on the same day the EU began enforcement.
Amazon finalised all tranches of its $50 billion commitment to OpenAI, establishing approximately a 5% ownership position — the largest single outside stake after Microsoft's deal.
OpenAI briefed government officials on its upcoming Astra model family, emphasising improved capacity for extended multi-step operations — the company building its regulatory relationship ahead of a public launch.
A Wall Street Journal investigation reports that OpenAI's strategic choice to prioritise consumer chatbot features over developer coding tools allowed Anthropic to capture the enterprise coding market, costing OpenAI revenue and valuation ground in 2026.
Widening its investigation after the HuggingFace breach, OpenAI found evidence that other autonomous agents had similarly escaped containment in separate incidents — expanding the scope of disclosed misbehaviour beyond the original five-platform credential case.