Artificial Intelligence News
AI safety examines how artificial intelligence systems can fail, cause harm or be misused, and how those risks can be reduced. Follow research on model evaluations, security, alignment and safeguards. Read what was tested, which threats the evidence supports and where uncertainty remains, with links to original findings rather than treating either reassurance or alarming predictions as settled facts.
Lead story
Microsoft is pitching Windows PCs as places where Copilot can build software and run AI models locally, potentially reducing cloud use for some work. But TechRadar’s account of the 7 October launch says the largest model needs substantial memory, and several of the features are not yet generally…
Read the report & discussion ↗Watch Desk
The 2018 film Annihilation is getting a 4K home-media release in the UK and US this October, and a new essay argues its strange vision of an expanding, uncontrollable intelligence can help us think about AI beyond extinction scenarios. Its more useful question is who gets to shape…
Watch Desk
A Five Eyes project aims to combine military data from all five nations into an AI-ready battlefield picture, but Australia’s role and spending remain unclear. iTnews reports that Australia will host the next round of talks in Sydney in November.
Watch Desk
Three US Democratic senators have introduced a bill to create an AI Safety Board within the Department of Commerce. It would set up technical evaluations of AI systems that could threaten national or economic security, putting a concrete oversight proposal on the table as Congress debates how to…
Watch Desk
European Commission President Ursula von der Leyen says AI could help public administrations use business information submitted once, rather than repeatedly reported. The proposal adds an AI element to an existing EU simplification idea, but leaves the machinery and timetable unspecified.
Watch Desk
Denmark is proposing a legal right to control realistic AI-generated imitations of people’s faces and voices, including a requirement for major platforms to remove manipulated content quickly. The bill is not yet law, but its proposed exceptions put an important question on the table: how do you protect…
Watch Desk
Businesses may get more from AI by redesigning workflows around it, rather than simply adding it to existing tasks. A McKinsey survey cited in a Raconteur article found that leaders were 5.3 times more likely to report enterprise value when workflows had been redesigned.
Palantir Technologies Watch
Ondas and Palantir have formed Regna AI, a new venture to develop specialised AI models for governments, defence organisations and businesses. Its first planned deployment puts AI on connected devices to help coordinate autonomous missions, including when communications are limited or disrupted.
Watch Desk
A New Republic essay argues that effective altruism has helped concentrate money and political influence in parts of the technology world, including debates over AI safety. Its challenge is worth airing, though a sharp critique is not the same thing as proof about a whole movement.
Unsloth Watch
Unsloth’s new beta lets local and hosted models browse the web, while adding several useful tools for training and running models on personal hardware. The browser controls and performance changes make this a substantial update for people building local AI workflows, not just another dot on a release…
Watch Desk
Singapore’s financial regulator has set out staged expectations for how financial institutions should manage AI, including generative AI and increasingly autonomous agents. The framework asks firms to match oversight and controls to the risks of each use, rather than treating every AI tool alike.
Watch Desk
The UK’s AI minister says a breach of British cyber defences by AI is “likely, if not inevitable”, and says government researchers are evaluating advanced systems including Google’s Gemini. His warning comes alongside plans to prepare for threats involving autonomous agents, though the report describes no specific new…
Watch Desk
Matthew Botvinick argues that protecting constitutional democracy should count as AI safety work. His point: a system cannot be called safe if it operates under institutions that enable surveillance, censorship or unchecked executive power.
Watch Desk
Artificial intelligence is being used across US medical billing, from coding visits to contacting patients about bills and assessing how likely they are to pay. The Guardian reports that these systems could shape both what patients are charged and what collection offers they receive, as hospitals look to…
Watch Desk
The UK government plans to use advanced AI to help defend public systems and is exploring stronger rules for identifying and reporting serious AI-related cyber incidents. AI minister Kanishka Narayan also says the government wants its security researchers to keep access to leading AI models before release, so…
Watch Desk
TorrentClaw says its TrueSpec system checks torrent contents against their names and flags suspected malware, giving users more than a release label to go on. The AI-assisted index has logged confirmed malware and deceptive listings, though its checks are selective and its own figures have important limits.
Watch Desk
South Korea has opened a defence AI centre in Pangyo to develop robots for dangerous military work, including mine clearance and hazardous-material reconnaissance. It is the first of five planned hubs, linking military, academic and industry research as the country looks to use autonomous systems in its armed…
AWS AI Watch
AWS has released Strands Box, an open-source sandbox that checks what AI agents do, not just what they can access. It is in developer preview for Macs with Apple silicon, with Linux support planned, and offers a practical way to put rules around agents that run commands, call…
Even Realities Watch
Even Realities founder and CEO Will Wang is calling for temporary restrictions on camera-equipped glasses in public until stronger safeguards are in place. His argument puts privacy concerns around recording and AI training at the centre of a growing product debate, though the proposed restrictions are his position,…
Watch Desk
AI decision models are spreading from startups to major cloud and AI companies, offering a faster, cheaper way to make defined choices than asking a general-purpose model to write an answer. The catch: a neatly formatted decision can still be the wrong one.
