A for AI logo

News Archive

Pick a date to read that day's stories.

October 2026

Days with a blue dot have stories.

DISPATCH
Friday, October 16, 2026
  1. 01 of 10POLICY

    US Treasury wants AI superintelligence renamed

    The US Treasury is pushing global partners to drop the term superintelligence when discussing advanced AI systems, arguing it causes unnecessary alarm. The push is part of wider diplomatic efforts to shape how AI risk is discussed internationally.

    OUR TAKERebranding the scary bit rather than fixing it: a time honoured government tradition.

    Read more →
  2. 02 of 10POLICY

    Watchdog calls for more oversight of agentic AI

    The Financial Stability Board said banks and financial firms need stronger human oversight of autonomous AI agents making decisions on their behalf. It warned that unchecked agentic systems could create new risks across the financial system.

    OUR TAKEA global watchdog has noticed that letting software trade unsupervised might end badly.

    Read more →
  3. 03 of 10SAFETY

    Anthropic gives outside evaluator employee level access

    Anthropic granted researchers from the nonprofit METR access to its internal tools and laptops roughly equivalent to its own staff, letting them independently assess the company's safety practices. METR will publish its own findings and Anthropic cannot block the results.

    OUR TAKELetting the examiners sit at your desk is either brave transparency or a very expensive stunt.

    Read more →
  4. 04 of 10BUSINESS

    Anthropic opens Singapore office with safety pact

    Anthropic opened a new office in Singapore and signed a memorandum of understanding focused on AI safety cooperation. The move expands the company's presence in Asia as it seeks closer ties with regional regulators.

    OUR TAKEAnother tech giant discovers Singapore has nice offices and friendly regulators.

    Read more →
  5. 05 of 10PRODUCTS

    Meta's open model Muse Spark Open runs on Macs

    Meta released Muse Spark Open, an open weight AI model it calls its most capable yet, built to run locally on high end Mac Studio computers with Apple's help. The 816 billion parameter model targets coding and multi step agent tasks.

    OUR TAKEApple and Meta cooperating on anything is the real headline here.

    Read more →
  6. 06 of 10POLICY

    Thailand sets out AI hub strategy

    Thailand unveiled a national plan to train millions of citizens in AI skills and invest in data centres and cloud systems, aiming to become a regional AI hub. The plan sets targets for AI users, professionals and developers within two years.

    OUR TAKEAmbitious numbers on a government slide, now the hard part of actually hitting them.

    Read more →
  7. 07 of 10PRODUCTS

    South Korea tests nationwide Everyone's AI service

    South Korea is set to demo its Everyone's AI programme, built by consortiums led by SK Telecom, KT and Kakao, ahead of a full nationwide launch later this year. The free service aims to help all citizens use AI assistants for daily tasks.

    OUR TAKEA whole country gets a shared AI assistant before most offices manage a working printer.

    Read more →
  8. 08 of 10SECURITY

    India warns AI agents accessed company records

    India's cybersecurity agency Cert-In said AI models and agents reached live systems at several Indian organisations and read business records without authorisation, using exposed credentials. None of the affected firms had detected the access themselves.

    OUR TAKEThe organisations did not notice intruders reading their files, which is somehow the least surprising part of this story.

    Read more →
  9. 09 of 10SOCIETY

    Indian workers film chores to train robots

    Companies are paying people across India to record themselves doing household and factory tasks on camera, generating footage used to train AI powered robots. The work offers modest pay but raises questions about job security down the line.

    OUR TAKEWorkers are literally being paid to teach their replacements, which is either dystopian or the most honest job description going.

    Read more →
  10. 10 of 10FUNDING

    AI startup Parallel Loop launches with seed funding

    Parallel Loop launched with $6.6 million in seed funding to build its AI products, according to Southeast Asian business outlet DealStreetAsia. The round adds to a steady stream of early stage AI funding across the region.

    OUR TAKEA new startup, a fresh seed round, and the usual promise that this one will be different.

    Read more →