Anthropic files for the stock market
Anthropic, the company behind Claude, has filed its IPO paperwork and updated it with third-quarter figures on 2 November. The filing shows USD 17.0 billion in revenue for the quarter and USD 33.2 billion for the first nine months of 2026, alongside heavy spending on computing capacity (SEC filing; CNBC).
Why it matters: this would be the first big public listing from a frontier AI company, and the filing is a rare look at the real money behind the tools Malaysian offices use every day. Anyone planning training budgets or tool subscriptions gets a clearer picture of how the industry is funded.
Caveat: a filing is a snapshot, not a settled outcome; the listing and its valuation are still ahead.
Illustrative photo: NYSE trading floor, not Anthropic premises. Carol M. Highsmith / Library of Congress, via Wikimedia Commons. No known restrictions. Cropped and resized.
OpenAI ships GPT-6.1 Terra for work tasks
OpenAI released GPT-6.1 Terra on 26 October. It is available in the API and in ChatGPT Work and Codex, not in the standard ChatGPT chat, and OpenAI's own release notes list the pricing from USD 1 per million input tokens (OpenAI Model Release Notes).
Why it matters: new models keep landing in the work tiers first, so the gap between free and paid tools keeps widening. Teams that standardise on one setup should re-test it against new releases rather than assume last quarter's choice still holds.
Caveat: availability and pricing differ by plan and region; check what a Malaysian workspace actually has before promising anyone a feature.
Photo: Pioneer Building, San Francisco, photographed in 2019. HaeB, via Wikimedia Commons, CC BY-SA 4.0. Resized; display crop.
Meta rebuilds how it picks which ads you see
Meta announced three changes to its advertising system on 28 October, including a new ranking layer built on large language models and upgrades to its Andromeda and GEM systems. Meta's ads infrastructure leadership says the system now scores over a billion ads per second, and credits recent ranking work with more clicks and conversions on Facebook (AdExchanger).
Why it matters: anyone running Facebook or Instagram ads in Malaysia is buying from a system that changed underneath them. New ads may find their audience faster, but results from old campaigns are not a reliable guide to new ones.
Caveat: the performance figures are Meta's own; treat them as direction, not as a promise for any single account.
Photo: Nokia621, via Wikimedia Commons, CC BY-SA 4.0. Resized; display crop.
Instagram tests letting you tune your own algorithm
Instagram is testing a "Your Algorithm" feature that lets users add or remove topics in their Reels feed, with Explore and Threads to follow, according to a 27 October report citing chief executive Adam Mosseri (My Electric Sparks). This is second-hand reporting, not an official announcement, and there is no public release date. Facebook launched a similar topic-control tool for Feed and Reels on 14 October.
Why it matters: reach on Instagram may soon depend partly on topics viewers choose for themselves, not only on what the system guesses. Malaysian businesses and creators who post Reels should expect audience tastes to become more visible and more adjustable.
Caveat: this is a test, not a shipped feature; controls can change or never roll out widely.
Illustrative photo of Instagram in 2016, not the new feature. Santeri Viinamäki, via Wikimedia Commons, CC BY-SA 4.0. Cropped and resized.
Malaysia drafts an AI Governance Bill
The Digital Ministry is drafting an AI Governance Bill covering safety and accountability, including identity impersonation and false information. Minister Gobind Singh Deo told the Dewan Rakyat on 28 October that supporting measures such as MY-AISafe, an AI Code of Ethics and a Sovereign AI Cloud under Budget 2027 are in progress (The Edge Malaysia).
Why it matters: rules for AI use in Malaysian workplaces are moving from guidance to law. Companies using AI for content, customer data or hiring should start keeping clear records of how and where they use it.
Caveat: this is a bill being drafted, not yet law; obligations will depend on the final text.
Photo: CEphoto, Uwe Aranas, via Wikimedia Commons, CC BY-SA 3.0. Resized from the original.
329,000 Malaysian teachers trained on AI
More than 329,000 teachers have received AI training, and AI literacy and basics will enter the school curriculum from next year, Education Minister Fadhlina Sidek told Parliament on 29 October. The ministry's AI Literacy Guidelines have been in force since May (The Star).
Why it matters: AI literacy is becoming a national baseline, starting in schools. Employers will meet a generation that treats AI as an ordinary tool, which raises the bar for workplace training rather than lowering it.
Caveat: training delivered is not the same as skills absorbed; schools and workplaces will both need follow-through.
Photo: CEphoto, Uwe Aranas, via Wikimedia Commons, CC BY-SA 3.0. Resized from the original.