Lead story / products

Airbnb targets the first half of 2027 for an AI travel agent

Airbnb plans to begin rolling out its own AI travel agent to guests in the first half of 2027. In its November 5 shareholder update, the company outlined an experience built around travel rather than a standard chatbot: it would assemble maps, photos, comparisons, calendars and itineraries in real time to match a guest’s question.

The design reflects Airbnb’s view that choosing a trip is not just a chore to outsource. The company says 80% of travelers enjoy researching trips and its bookings average three guests. It wants the agent to help people compare options and decide together, preserving the browsing that already happens through maps, listing photos and reviews. Those behavioral figures are not performance results for the forthcoming agent.

The other part of Airbnb’s pitch comes after the choice. Because it handles reservations, payments and customer support, the company says its agent could help with changes, payment problems, refunds, rebooking and 24-hour safety support. That access to the underlying booking is its argument for offering something an outside assistant cannot easily provide on its own.

Airbnb’s longer-term ambition is to help guests plan an entire trip in one place, spanning homes, hotels, services and experiences. That broader reach is an ambition, not the promised scope of the first release. The first-half 2027 window is a target for starting the guest rollout, not a promise that the whole vision arrives at once.

launch

Cursor lets users build hosted web apps with sign-in from a prompt

Cursor’s Build Live Apps combines app generation, hosting, sign-in and optional Postgres databases on paid plans with cloud agents enabled. Users must explicitly publish; security checks look for common issues such as exposed credentials but do not guarantee a vulnerability-free app. Team administrators can add review rules, authorize blocked releases and take apps offline.

security risk

Prosecutors charge three over an alleged $129 million AI impersonation scheme

Federal prosecutors in New York announced indictments against three people over an alleged $129 million AI-assisted scheme. Court documents describe emails and calls mimicking CEOs, followed by a fake lawyer’s instructions to wire payments. Prosecutors say thousands of companies were targeted over more than 18 months; that is not a count of firms that paid. The allegations remain unproven.

launch

Perplexity launches AI agents for ongoing jobs, limited to Max subscribers

Perplexity says its Teammates retain project context, run in the background and can involve people or other agents, including contacting coworkers through Slack. Access is limited to Max subscribers on the web, with other platforms promised but no dates. Usage draws on subscription credits after a promotional grant whose size and expiry are undisclosed, leaving the launch allowance unclear.

Also worth knowing

•  OpenAI’s GPT-6 Astra leads a benchmark built around bluffing and spotting lies

•  Discovered Materials offers $1 million for five unsolved science problems

•  OpenAI releases an AI security model for verified defenders

•  Investors question Anthropic’s reported $1.5 trillion IPO valuation

•  Thinking Machines releases a coding model but holds back downloadable weights

•  James Gates sues OpenAI, alleging ChatGPT helped plan his poisoning

•  OpenAI finds its model did less work on tasks labeled safety research

•  Claude can analyze linked bank accounts but cannot move money

•  Google cuts AI video prices 30% and extends clips to 15 seconds

•  Matt Husk says AI labs sought collaborations after his chatbot takedowns

Daily tool drop

6 AI tools worth knowing today

The Internet Had a Point

Astra Billion-Dollar Prompt

Reader check-in

Help shape tomorrow's briefing

One click tells us what to keep, improve, or tighten.