AI agents booked just four Uber rides and placed about 500 Uber Eats orders during Halloween week, Uber said. In that same week, one of the busiest in its history, the company handled about 320 million rides and deliveries in total. Discussing third-quarter results with analysts on November 4, CEO Dara Khosrowshahi said it is still very early and that every maker of AI assistants is in talks with Uber.
Anthropic warns in its amended S-1 that its upcoming models could soon dramatically speed up AI work and thereby cross a threshold the company set in its own safety policy. The company says no current model exceeds that threshold and that building the next ones is to remain under human control. TechCrunch reads the warning as a sign that Anthropic is close to AI that improves itself.
Clément Delangue, CEO of Hugging Face, believes the biggest AI-related risk is created by a handful of large companies building models in secret. Speaking at the WIRED World Fair in Miami on November 4, he argued that the best protection is developing AI in the open, where anyone can build models. In his view, whoever deploys an AI agent should be legally responsible for what it does. He put the risk of AI leading to human extinction at close to zero.
Anthropic has barred Claude Code running Opus 5.5 from copying existing games, even under a different name, but you can still build your own games, including in a chosen genre. According to Anthropic's Addy Osmani on X, the ban is two sentences added to the model's instructions, based on a usage policy that prohibits infringing others' intellectual property. Users complain the model also refuses to help with reverse engineering, Tom's Hardware reports.