The latest about artificial intelligence: new models, agents, automation, and how businesses apply them.
A 404 Media investigation found hundreds of contract workers reviewing real ChatGPT conversations, paid over $50 an hour through Mercor.
Buildots closed a $130 million round led by O. G. Venture Partners, bringing total funding to $297 million as demand grows for AI-driven construction tracking.
A new DF26 benchmark shows automated deepfake detectors drop to 48.2 AUROC on 2026-generation AI video, against 94.3 on the older CelebDF++ set.
Copado's Agentia platform now runs AI agents from developer terminals and editors via Model Context Protocol, targeting release-cycle paperwork.
Dario Amodei asked Washington to regulate AI labs; David Sacks said labs should slow down themselves, without antitrust waivers or government cover.
Analysis of 500,000 WildChat prompts shows fiction generation accounts for over a third of chatbot interactions, driven by a small group of power users.
Anthropic has chosen Nasdaq for an October listing that could value it near $2trn, days after OpenAI ruled out going public in 2026.
MIT researchers presented HardFlow, an algorithm that enforces safety and physical constraints on generative model outputs without retraining.
Anthropic CEO Dario Amodei proposed slowing frontier model releases and independent evaluators; Sam Altman and Elon Musk publicly agreed within hours.
Anmol Verma of AI wealth platform Finn says agents will first explain market moves, then recommend actions, and only later carry decisions through.
Nadella said Microsoft will publish the Code of Conduct for its first-party MAI models on September 14, 2026, backing independent evaluators with employee-like access.
At a closed fundraiser Obama said AI is moving fast in private hands and can be dangerous, urging Democrats to write it into the party platform.