Sunday 19 July 2026

The Brief – 19/07/2026

The top things worth knowing about in AI today.

  1. US firms switch to cheaper Chinese models

    DoorDash, Airbnb, Siemens and coding startup Cursor are all now running production workloads on Chinese open-weight models from Moonshot AI and DeepSeek. A UBS analysis puts those models at roughly 95 percent of US capability for about 10 percent of the cost, and Chinese open-source models accounted for 41 percent of Hugging Face downloads in March. If your AI bill is climbing, the question is no longer whether a cheaper model can do the job, but which tasks actually need the frontier.

    Read more
  2. Meta sued over AI-assisted layoff selection

    Twenty-six former Meta employees filed suit in federal court in Oakland, alleging the company used activity monitoring, AI token-usage dashboards and algorithmic performance rankings to pick roughly 8,000 people for its May layoffs. Eight plaintiffs had taken pregnancy-related leave and four had recently been on parental leave. They want an independent audit of the selection process, which would be the first real test of whether behavioural telemetry can lawfully drive redundancy decisions.

    Read more
  3. New York halts new hyperscale data centres

    Governor Kathy Hochul signed Executive Order 62, pausing state environmental permits for any data centre project at 50 megawatts or above for up to a year while a regulatory framework is written. It is the first statewide moratorium in the US and will assess energy demand, water use and air quality. Within 60 days the state will issue guidance for councils negotiating community benefits with operators.

    Read more
  4. Anthropic ties Australian spend to copyright clarity

    Anthropic has told Treasurer Jim Chalmers that its planned A$21.6 billion Australian investment, covering 1.4 gigawatts of data centre capacity, depends on Canberra clarifying local copyright obligations. The company is not seeking an exemption, which the Albanese government has already ruled out, but wants certainty on its liability to rights holders. It sets a direct price on how quickly Australia resolves the training-data question.

    Read more
  5. Gemini 3.5 Pro misses another launch date

    Google let the 17 July target for Gemini 3.5 Pro pass without a release, the third slipped deadline after engineers scrapped the original base model and restarted pretraining over problems with recursive tool-calling. Only Gemini 3.5 Flash is generally available; Pro remains in limited preview for select Vertex AI customers. Reports suggest Google is weighing a stopgap Flash release instead. Plan capacity on what has shipped, not what is promised.

    Read more
  6. Two hundred economists urge action on AI

    More than 200 economists and AI researchers, including 16 Nobel laureates, signed an open letter coordinated by Stanford's digital economy lab warning that AI could reshape the economy faster than the Industrial Revolution did. They are calling for incentives, guardrails and institutions to be built now rather than retrofitted. Lead author Anton Korinek put it plainly: waiting for certainty means arriving too late.

    Read more
  7. AI vendors spend billions on deployment teams

    Microsoft has committed $2.5 billion and about 6,000 staff to a new Frontier Company that embeds engineers inside client organisations. Amazon has put up $1 billion, OpenAI raised over $4 billion for its deployment arm, and Anthropic launched a $1.5 billion joint venture. Job listings for forward-deployed engineers rose more than 800 percent across 2025. The labs have concluded that selling a model is not the same as making it work.

    Read more
  8. Enterprise AI stalls on readiness, not capability

    Fifty-seven percent of enterprises now have AI embedded in core processes, up from 35 percent a year ago, but only 32 percent have hit even one of their top two AI objectives. Among companies above $1 billion in revenue, 71 percent of executives named organisational readiness as the main barrier and just 11 percent blamed the technology. Only 19 percent of workers say they feel confident using the tools. The gap is a change-management problem wearing a technology costume.

    Read more
The Weekly AI Brief

Practical AI, distilled.

A short read every week — the few things worth your time, and nothing that isn’t.