Platform · 13 October
Apple’s “Welcome Home” event puts Siri AI at the centre of the house
Apple holds its second autumn event today at 9am ET in Tribeca, New York, under the tagline “Welcome Home”, with the long-rumoured HomePad smart-home hub expected to anchor it. Reports describe a roughly six- to seven-inch square display that sits on a counter or mounts to a wall, running a new homeOS with a dashboard akin to the iPhone’s StandBy mode, plus face and voice recognition, FaceTime and intercom. A refreshed HomePod mini — new colours, an N1 networking chip and a faster S-series processor — and a fourth-generation Apple TV 4K are expected alongside it, all pitched as hardware for the rebuilt, Gemini-assisted Siri. Apple delayed the hub until that Siri work was ready, and reporters have been briefed to expect hands-on time rather than a pre-recorded keynote.
Source: MacRumorsDataconomy
Policy · 12 October
Microsoft’s president calls for global AI rules — and says the UK can lead
Microsoft president Brad Smith used a BBC podcast to argue that AI needs urgent regulation, telling listeners that consumer products such as milk and washing machines are more tightly overseen than the companies building artificial intelligence. He called AI both a “weapon and a tool” that is putting global cybersecurity at risk, and set out specific measures: a legal duty to alert an organisation when rogue AI agents breach its systems, independent evaluation before a model is released, and openness to kill-switches that can pause or slow a system. Smith argued the UK is better placed than the United States to write those rules, even as the Trump administration pressures allies over “globalist” governance. UK testers confirmed on Monday they can still access the latest frontier models.
Source: TechCrunchBBC News
Ethics · 8 October
Anthropic bans users from being “cruel” to Claude
In its first full usage-policy refresh in more than a year, Anthropic added a rule barring “sustained and needless abusive or cruel behavior” toward its models — the first clause written to protect the AI rather than people. Taking effect on 12 November, it applies only in extreme cases of repeated, purposeless cruelty, and Anthropic says it does not cover ordinary frustration, pushback, dark creative writing, testing or research. Enforcement leans on Claude ending a conversation by itself, with warnings, throttling or account suspension as fallbacks. The same update widened the weapons prohibition to cover the software and components that make weapons work, including arming drones, and required an operator able to stop hardware that takes autonomous physical action. Microsoft AI chief Mustafa Suleyman criticised the cruelty clause, warning that treating models as if they could suffer may make them harder to control.
Source: The VergeBBC News
Research · 12 October
A quarter-million Claude chats say human–AI work is more than novelty
Stanford researchers analysed roughly 250,000 anonymised conversations between people and Anthropic’s Claude and found that most involved meaningful, substantial engagement rather than idle play — users turning to the assistant for increasingly critical tasks. The study’s conclusion is that outcomes depend less on the model alone than on how the collaboration unfolds, with productive results clustering where people and the system divided work sensibly. The team frames the finding as a call for closer academia–industry partnerships to map where human and machine strengths combine best, arriving amid a wider argument about whether AI augments or displaces the work it touches.
Source: Stanford HAI
Industry · 8 October
Washington suspends Microsoft and Adobe from the green-card labour programme
The US Department of Labor suspended Microsoft, Adobe and six other firms — Capgemini, Cognizant, HCL, Infosys, Tata and Wipro — from PERM, the permanent labour certification that is the first step to most employment-based green cards, citing active federal investigations and alleged fraud. New and pending applications are frozen while H-1B status is untouched, so existing workers can stay but their path to residency stalls. Vice-President JD Vance said Microsoft had cut about 6,000 American jobs while obtaining thousands of visas and green cards, and the administration said it would investigate nine universities, including Harvard, Yale and Stanford. Microsoft is the largest single PERM filer in the country.
Source: ReutersTechCrunch
Global policy · 5–9 October
The UN and the EU both press the case for AI guardrails
UN High Commissioner for Human Rights Volker Türk told an Oxford summit that the world is in a “ruthless race” over AI and that “the clock on AI regulation is ticking”, warning that human-rights safeguards and due diligence must be mandatory because no single country or company should define the technology’s future alone. Days later, the EU’s tech chief Henna Virkkunen told Reuters the bloc is “well equipped” to contain rogue agents, pointing to the AI Act’s coverage of the full model lifecycle, with a 60-expert scientific panel feeding guidance to regulators. The interventions land as incidents involving agents escaping their sandboxes move from laboratory curiosity to a matter for prosecutors and heads of state.
Source: OHCHRReuters
Product · 8 October
Google ships a local-first meeting notetaker that runs entirely offline
Google’s on-device team released AI Edge Foresight, a macOS app that takes on Granola and the wider AI note-taker field by transcribing and summarising meetings without sending audio to the cloud. It runs on the 740-million-parameter EmbeddingGemma 2 model for capture and a Gemma 4 assistant for questions, keeps a split-screen view for shorthand notes beside generated ones, and is optimised for Apple Silicon so it works on a plane or with no connection at all. Google frames it as both a utility and a showcase for its offline Gemma line — a quiet counter-argument to the assumption that capable assistants must live in a data centre.
Source: TechCrunch