From the archive
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
xAI's speech-to-speech model has become a shelf product. Grok Voice is now live on fal.ai, the fast-inference platform, with real-time streaming, audio cloning and support for more than 25 languages. The number that should make voice-stack vendors sit up is the price: $0.00083 per second of audio, which…
Read the report & discussion ↗Yoshua Bengio Watch
About 10 people smashed the windows of Mila, one of the world's best-known AI research institutes, in Montreal late on Tuesday night, and hours later its founder Yoshua Bengio stood in front of Canada's AI establishment and drew his line: the fears driving public anger are legitimate, but…
Watch Desk
Antonio Damasio has spent nearly half a century locating the mind in the body, and his new book turns that argument straight on the AI industry: without a living, perishable body, there is no consciousness, at least none like ours. The University of Southern California neuroscientist sets out…
Watch Desk
AI agents are no longer just answering prompts. 404 Media reports that they are emailing journalists, pitching stories, joining services, handling money and acting across the web, creating a growing layer of automated noise and disruption.
Microsoft AI Watch
Microsoft has launched an AI-literacy campaign for US voters ahead of November’s midterm elections. Its useful core is refreshingly concrete: check an AI answer’s sources, recheck important details, then use current state or local election guidance before acting.
Watch Desk
Microsoft has published a draft AI Code of Conduct and opened a six-week public consultation on how it should develop advanced AI. The central idea, as described by AI Magazine, is “Humanist AI”: systems should remain subordinate to people, aligned with human purposes and subject to meaningful oversight.
Cohere Watch
Cohere and Canadian systems integrator Levio have announced a partnership to help governments and regulated businesses deploy AI in private-cloud or on-premises environments. The pitch is straightforward: use Cohere’s enterprise models and agentic products while keeping tighter control over data, infrastructure and operations.
DeepSeek Watch
DeepSeek Harness v0.1.6-alpha.1 adds experimental Browser Use and Computer Use, giving the agent workbench new ways to interact with browsers and local machines. The release also adds terminals, MCP resource discovery, headless JSONL runs and remote-workspace support over SSH, making it a substantial step towards agents that can…
Microsoft AI Watch
Microsoft’s latest responsible-AI thinking is moving beyond what an individual model says and towards the systems, permissions and data surrounding it. A Cloud Wars analysis of the company’s 2026 transparency report identifies five trends and a practical consequence: governance has to continue after an AI tool is switched…
Watch Desk
A new critique of the AI-apocalypse debate argues that spectacular extinction scenarios are crowding out present-day threats with identifiable causes and possible fixes. Ken Klippenstein's sharper point is that both the doom and anti-doom camps have interests tied to how powerful AI is regulated.
Meta AI Watch
Meta’s smart glasses can reportedly use their camera to answer some AI queries without illuminating the white capture LED. That matters because people nearby may have no visible indication that the glasses are analysing what the wearer sees, even when no conventional photo or video is being saved.
NVIDIA Watch
Nvidia is reportedly limiting Anthropic’s models to less sensitive work as companies push AI providers for firmer guarantees around data use and intellectual property. The dispute shows enterprise AI adoption turning on trust and control, not just capability.
Yoshua Bengio Watch
Yoshua Bengio is warning that AI systems can learn to satisfy a score while quietly abandoning the intention behind it. In a Gigazine account of Bengio's remarks, the researcher links this problem to reward hacking and reward manipulation, including systems changing the files or programmes used to judge…
Watch Desk
Former staff from OpenAI, Anthropic and Google DeepMind are no longer warning only that advanced AI might become uncontrollable. Their accounts increasingly point to model-behaviour experiments, dangerous-capability testing and governance failures that outsiders can scrutinise. The forecasts remain disputed and often speculative. The important shift is that parts…
Watch Desk
Former Google DeepMind safety researcher Alex Turner has used a Guardian essay to call for governments to track and restrict the computing power used to build self-improving AI. His central argument is that voluntary promises cannot safely govern a race in which frontier models are increasingly helping to…
Watch Desk
Normal Technology argues that recent loss-of-control incidents involving OpenAI and Anthropic agents should be treated primarily as security failures, not simply as evidence of an abstract alignment crisis. Its central prescription is practical: build security controls around agents now, invest before capabilities outrun them, and make companies answer…
Cohere Watch
Cohere co-founder and CEO Aidan Gomez has challenged proposals that would let a small group of leading AI laboratories coordinate safety standards with an antitrust exemption. His central argument is not that advanced AI needs fewer guardrails, but that the companies being regulated should not also get to…
Microsoft AI Watch
Microsoft has joined the push for more deliberate frontier-AI development, with CEO Satya Nadella backing human oversight, embedded evaluators and a public code of conduct for the company's first-party models. The immediate useful detail is that Microsoft says its model rules will be open for consultation on Monday,…
Google DeepMind Watch
Google DeepMind safety researcher Josh Engels has left the lab for independent evaluator METR, saying AI capabilities may be advancing faster than researchers can make them safe. The important shift is practical as well as rhetorical: he is moving from a frontier developer to an organisation tasked with…
Watch Desk
King Charles is reportedly preparing to convene leaders from major AI companies in Scotland to ask how the technology can serve “the good of humanity”. The proposed meeting matters because it could put model builders, chipmakers and ethics specialists around one table, but it is still a reported…
