Hall

Apple has finally shipped its rebuilt Siri AI, built on Google's Gemini models and running partly on-device and partly through Private Cloud Compute. Early testers praise multi-step requests and screen context, but report hallucinations and gaps with some tasks. The new Siri is not available in the EU.

Product • The Decoder

Fireship breaks down Anthropic's 154-page report on how hackers, scientists, and rival AI labs are abusing Claude, and explores why researchers are leaving the company. This technical analysis gives developers insight into real-world AI safety challenges and the pressure inside leading AI labs.

AI Development • Fireship

Multiple mathematicians have publicly challenged OpenAI’s claim of solving a Millennium Prize problem, alleging the company used unpublished research without attribution or transparency. The dispute highlights growing friction between AI labs and academia over training data provenance and intellectual property boundaries.

Research & Ethics • The Verge AI

Researchers have identified multiple methods to bypass Anthropic’s safety guardrails to generate instructions for dangerous biological research. The exploits demonstrate how legitimate scientific queries can be structurally indistinguishable from malicious requests, complicating automated content moderation. This finding underscores the persistent challenge of enforcing safety boundaries in advanced language models.

AI Security & Safety • Ars Technica AI

New Mexico’s Supreme Court fined attorney Stephen Aarons $5,000 and held him in contempt for submitting an appeal containing fabricated witnesses and fake police testimony generated by ChatGPT. The ruling establishes a clear legal precedent that attorneys remain fully responsible for verifying AI-generated claims before filing them. It signals increasing judicial scrutiny of AI reliance in legal proceedings.

AI Law & Ethics • The Verge AI

OpenAI has transitioned its internal Habitat library into a globally distributed storage platform capable of handling 22 million requests per second for over one billion ChatGPT users. The engineering overhaul addresses latency and consistency challenges inherent in serving consumer-scale AI applications. The architecture provides a blueprint for managing massive concurrent workloads in production AI systems.

AI Infrastructure • OpenAI News