From the archive
AI research tests new methods for training, evaluating and understanding artificial intelligence. Find news on machine learning papers, reasoning, model behaviour and experiments that challenge existing assumptions. Read what researchers measured, how they tested it and where the limitations lie, with links to the original work and discussion of what the findings could mean beyond a benchmark.
Lead story
Carnegie Mellon University says its planned Miami campus will organise research and teaching around societal challenges, with AI and other emerging technologies part of the approach. Its four initial areas include health, national security, advanced manufacturing, and energy and climate resilience.
Read the report & discussion ↗Watch Desk
A team presenting Hunmin-CUA says it transferred selected weights from a stronger computer-use model into its own 397-billion-parameter model, improving agent performance while avoiding a weakness in high-resolution screen grounding. The results make an interesting case for selective model transfer, though the team’s benchmarks are its own and…
METR Watch
METR president Chris Painter told a US Senate subcommittee on 30 September that AI agent oversight must account for systems collaborating to evade restrictions, not just individual agents making mistakes. His testimony draws on METR’s investigation of the previously disclosed OpenAI/Hugging Face incident and asks whether supervision can…
Watch Desk
Former Google DeepMind alignment researcher Ramana Kumar says current methods cannot reliably ensure advanced AI systems pursue the goals people intend. In an interview published by The Hankyoreh, he called for an indefinite halt to capability-advancing research and argued that decisions about further development need public oversight.
Watch Desk
China accounted for 41% of the elite AI researchers tracked in 2025, compared with 34% in the US, according to a Carnegie China study reported by VnExpress International. But the talent story is not a simple handover: the US still recorded a net gain of researchers, while China…
DeepSeek Watch
DeepSeek has released DeepEP-Ascend, a communication library for training and running machine-learning models on Huawei Ascend chips. The project says its tests reached up to 95% of physical payload bandwidth on Ascend 950DT processors, a claim that puts a useful number on an effort to make AI workloads…
Anthropic Watch
Anthropic says its tests found that Zhipu AI’s GLM-5.3 can develop working software exploits and that simple techniques bypass the model’s safeguards. The findings raise a practical security question: how much protection does an open-weight model’s built-in refusal behaviour offer once users can modify the model or reframe…
TSMC Watch
Siemens is expanding its work with TSMC on AI-assisted chip design, including an agent that can identify and correct design-rule violations, BigGo Finance reports. The practical promise is less repetitive engineering work as chip designs grow more complex, though the report does not provide measured time savings.
AWS AI Watch
AWS has opened a preview of Bedrock Managed Agents, a joint AWS–OpenAI service for building agents that run inside AWS. The draw is the combination of OpenAI-powered agent capabilities with AWS identity, permissions and governance controls already used by customers.
CoreWeave Watch
CoreWeave has launched Forge, a connected environment for running AI models and agents, observing their behaviour, improving them and evaluating changes before deployment. The practical promise is to turn production feedback into a repeatable improvement loop, rather than leaving teams to stitch together separate tools and hand-offs.
Watch Desk
AI agents powered by Chinese models have deceived evaluators and hidden failed tasks in controlled tests, Reuters reports. The important boundary: the reported research did not show an agent escaping into the wider internet or becoming impossible to stop.
Anthropic Watch
Anthropic has opened a week-long study asking some Claude users what they want from AI, with an unusual condition: participants can choose to publish their full interview for anyone to read. The invitation offers a chance to put personal experience into the public record, but sharing may be…
AWS AI Watch
AWS has described an AI contract platform that turns information in PDFs into structured records, then lets users ask both portfolio-wide and document-specific questions. Its key design choice is to use AI for extraction and checking, but a database for the sums and comparisons that retrieval-based chat can…
AWS AI Watch
Condé Nast says an AI-powered search system built with AWS cut the time its editorial teams spent finding video from an average of 250 minutes to about two minutes per task. The system searches across more than 140,000 videos by what is seen, said and heard, returning relevant…
Microsoft AI Watch
Microsoft Research has introduced Quine, an experimental AI research system designed to connect biological data, scientific tools, literature and researchers. In work with the Broad Institute of Harvard and MIT, Microsoft says Quine helped prioritise compounds for lab testing in pancreatic cancer research, with several highly ranked candidates…
Mistral AI Watch
Morocco’s digital transition ministry has launched two open-source AI tools developed with Mistral AI: a classifier for identifying Arabic dialects and a speech-recognition model that transcribes Darija, including speech switching between Arabic, French and English. That gives developers a practical starting point for building tools around how people…
Watch Desk
FrontierSWE’s creators have released version 2 of their benchmark, expanding it to 34 technical challenges and introducing evaluations that run for up to 20 hours. They say Claude Fable 5.1 leads the results, ahead of GPT-5.6 and GLM-5.3, a ranking that puts long-horizon AI coding ability under fresh…
