Morning edition — Thursday 12 November 2026

Google introduced the Gemini Decisions API on 11 November, powered by a new model, Gemini Decision 1 (gemini-decision-1). Instead of generating text, it scores a fixed set of options and returns probabilities and confidence scores, for tasks such as tool selection, policy checks, model routing and grading context.

One request carries a single input of text, images, audio, video or PDFs, up to 256K tokens, plus up to 1,000 independent questions: yes/no checks, choices among 2 to 255 options, or scores on a rubric. Google says it answers in 150 to 250 milliseconds end to end and charges $0.10 per million input tokens with no output charge; the shared input is billed once per request. It is available in Google AI Studio with a no-cost tier.

Google claims the model is nearly five times cheaper on its benchmark suite than Gemini 3.5 Flash-Lite. It enters a field where TypeSafe's Jev and OpenAI's Decisions API already sell similar decision models, VentureBeat notes.

United Airlines said on Wednesday 11 November that it is restricting access by unauthorised third-party AI agents and assistants, both on its own website and app and indirectly through travel agencies and aggregators. Its help page says any agent, bot, browser extension or browser feature that searches, books, buys, manages or collects information on united.com or in the United app without the airline's permission is unauthorised.

United says sharing account credentials with such assistants can expose personal data and lead to unintended purchases, account changes or use of MileagePlus miles, and that automated check-in could breach rules requiring a human to confirm a bag carries no banned items. Bookings an agent made before 11 November will not be cancelled, travellers can still use AI to plan trips without touching United's systems, and assistive technology for customers with disabilities is not affected. Travel agencies may use AI tools only with United's permission.

United says it does not yet work with any third-party AI agents. CNBC notes that companies are split: Amazon has blocked Meta's Muse agent while Walmart has partnered with it.

SpaceXSI launched Grok Bot 1.0 on 11 November, ending the beta of its always-on AI agents. Team Bots can be shared with people or groups and brought into Slack, Microsoft Teams and video calls, where they follow multi-person discussions and take part through the chat.

The company says the bots now offer to take on work unprompted: they monitor projects, follow up on unfinished tasks and bring people in when a decision needs human judgment. It also says they are faster and more reliable at browser work, documents and slide decks, and adds team settings and usage analytics. Team Bots are on Team and Enterprise plans; meetings are on Team, Enterprise and Ultra plans and must be switched on by Enterprise admins.

Santosh Janardhan, Meta's head of infrastructure and co-head of engineering, told staff on 11 November that he will step down, according to an internal memo first reported by Bloomberg and also obtained by CNBC. He has been at the company since 2009 and wrote that it is "time to pass the baton".

Dave Brown, who joined Meta this year after about two decades at Amazon Web Services, will succeed him. Janardhan will support the transition through the first quarter of 2027, and a Meta representative told TechCrunch he is moving into an advisory role on Meta Compute, the initiative that plans Meta's computing capacity for AI. The memo gave no reason for the change.