South Africa targets January 2027 for revised AI policy after earlier withdrawal ↗
South Africa is rebuilding its national AI policy after an earlier draft was pulled over fictitious, possibly AI-generated references. Not exactly the kind of “AI transformation” anyone wants.
An independent expert panel will review the document, repair bad citations, and help restore credibility. Two officials have been suspended while the issue is investigated - uncomfortable, but perhaps a warning label for governments leaning on AI.
DeepMind CEO: AI agents are a “practice run” for AGI ↗
Demis Hassabis says today’s agentic AI systems are, in effect, a rehearsal for AGI. That’s both exciting and a little thundercloud-in-a-teacup, depending on your mood.
His point: agents are becoming practical fast, and society is not preparing at the same speed. He wants governments, economists, and the public to treat this as more than a tech-sector hobbyhorse.
AI warfare is already here ↗
The military AI debate is no longer about sci-fi killer robots. It’s about targeting systems, surveillance tools, and companies trying - or failing - to set red lines.
Anthropic’s limits on autonomous targeting and domestic mass surveillance sit at the centre of the story, but the broader trend is blunt: militaries are already using AI to compress decision-making. The ethical guardrails look a bit like wet cardboard.
Sundar Pichai on AI, the future of search, and what’s happening to the web ↗
Sundar Pichai laid out how Google is remaking itself around AI, from Search to YouTube to Gemini-powered agents. The web, meanwhile, is being pulled into the blender.
He argues Google still wants a healthy open web, but AI answers and agentic search are changing the traffic game for publishers. A lovely promise, attached to a slightly terrifying machine.
Anthropic appoints KiYoung Choi as Representative Director of Korea ahead of Seoul office opening ↗
Anthropic is pushing harder into Korea, naming KiYoung Choi as its local representative director before opening its Seoul office. Korea is apparently one of Claude’s strongest markets relative to population size - a surprisingly spicy signal.
Choi comes from Snowflake and has led tech businesses across major cloud and software companies. The plan is enterprise partnerships, startup support, and closer ties with Korean research and government groups.
Anthropic’s Responsible Scaling Policy ↗
Anthropic updated its Responsible Scaling Policy with a tighter threshold for novel chemical and biological weapons risks. Tiny policy wording? Maybe. Big consequences? Also maybe.
The update also refines how Anthropic handles off-cycle risk updates for individual models. It’s safety paperwork, yes, but for frontier labs this stuff is basically the scaffolding under the rocket.
MAI-Image-2.5 launches at No. 3 on Arena ↗
Microsoft introduced MAI-Image-2.5, saying it ranks third on the Arena text-to-image leaderboard. The pitch is sharper instruction-following, better text rendering, and more coherent commercial-style visuals.
The practical bit is brand work: packaging, posters, product shots, layouts that don’t melt into noodle soup. It’s coming to Microsoft’s own playground and Foundry soon, or so it seems.
FAQ
Why is South Africa revising its national AI policy?
South Africa is revising its national AI policy after an earlier draft was withdrawn over fictitious references, which may have been AI-generated. The government is now relying on an independent expert panel to review the document, correct citation issues, and restore credibility. Two officials have also been suspended while the matter is investigated.
What did DeepMind’s CEO say about AI agents and AGI?
Demis Hassabis described today’s agentic AI systems as a kind of practice run for AGI. His concern is that AI agents are becoming valuable at speed, while governments, economists, and the wider public may not be preparing at the same pace. The point is less about publicity and more about readiness.
How is AI already being used in warfare?
AI warfare is already moving beyond science fiction and into practical military systems. The article points to targeting systems, surveillance tools, and faster decision-making as current areas of concern. Companies such as Anthropic are trying to define limits around autonomous targeting and domestic mass surveillance, but the broader trend is already underway.
How is Google changing Search with AI?
Google is reworking Search around AI answers, Gemini-powered agents, and more automated ways of finding information. Sundar Pichai says Google still wants a healthy open web, but AI search is changing how traffic flows to publishers. That creates tension between better answers for users and the future visibility of websites.
Why is Anthropic opening an office in Korea?
Anthropic is expanding in Korea by appointing KiYoung Choi as Representative Director ahead of its Seoul office opening. Korea is described as one of Claude’s strongest markets relative to population size. The company’s plans include enterprise partnerships, startup support, and closer work with Korean research and government groups.
What is Microsoft MAI-Image-2.5 designed to improve?
Microsoft says MAI-Image-2.5 is built to improve text-to-image generation, especially instruction-following, text rendering, and coherent commercial-style visuals. The practical focus is brand and production work, including packaging, posters, product shots, and layouts. Microsoft says it ranks third on the Arena text-to-image leaderboard and is coming to its own platforms.