Three labs confirm weeks of safety talks. Meta puts a price list on extra AI across its apps. Google ships a live voice model with per minute pricing.
In the news
OpenAI, Anthropic and Google DeepMind have been talking
Chris Lehane, OpenAI's global policy chief, told reporters on September 15. OpenAI has been working with Anthropic and Google DeepMind on AI safety for weeks. Bloomberg first reported the disclosure. TechCrunch covered it. Lehane is in Washington for catastrophic risk legislation. The talks follow Dario Amodei's Saturday essay on pacing the frontier. Altman, Hassabis and Musk supported the call. Altman said OpenAI would join Anthropic in embedding third party evaluators. Lehane said the firms do not need Amodei's proposed government waiver. 1
Meta One subscription layer
Meta announced Meta One on September 15 via Newsroom. The subscription service adds paid features and AI across Facebook, Instagram and WhatsApp. Plans start at $2.99 per month for single products. The Max plan is $499 per month. It offers the highest feature access and Meta Business Agent capacity. Meta says more than 50 features have launched. There are 15 million subscriptions and trials to date. Core apps and Meta AI stay free. 2
Gemini 3.8 Live voice API
Google announced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15. The models are native speech to speech Live API models. They accept text, image, audio and video input. They output text and audio. Pricing is $0.005 per minute for audio in. Audio out is $0.018 per minute. Google says Extended Thinking ranks first on Artificial Analysis Speech to Speech Quality Index at 82.6. Gemini 3.8 Live ranks second in Speech Agent Arena. The models are rolling out today on Gemini API and Google AI Studio. A private preview is available in Gemini Enterprise. 3
