These reports remain available for context. Their original source dates are shown below; an article’s release date does not make an older source new.
Technology · Latest source 2026-11-06
China's telecom regulator has issued a four-part control framework for phones, tablets and wearables with pre-installed agents. Developers must document compatibility; manufacturers must report major and over-the-air changes; and app platforms are told not to let agents force access without a developer declaration.
Government & Policy · Latest source 2026-11-06
A five-page consultation would require manufacturers of industrial, service and special-purpose robots made in China to register products, mark every unit before factory release and file the used code within 20 working days. It is not yet a final rule.
Health & Life Sciences · Latest source 2026-11-07
A peer-reviewed comparison gave five commercial language models the same 227 questions from one US Doctor of Pharmacy course. Overall scores clustered between 89.4% and 91.2%, while image questions and the highest Bloom category were materially weaker.
Security & Defence · Latest source 2026-11-06
Japan approved ¥10 billion for immediate cyber measures after repeated intrusions and data leaks. The package funds AI-assisted vulnerability checks and creates new response teams, but most capabilities still have to be built and independently tested.
Science & Research · Latest source 2026-11-06
A peer-reviewed framework estimates time-direction asymmetry in autoregressive generators without enumerating every history. GPT-2 and Qwen3-4B-Base assigned larger sentence-level irreversibility to causally ordered text in limited tests, while token-level results were dominated by syntax.
Security & Defence · Latest source 2026-11-06
Two NeurIPS-accepted studies attack the problem at different stages: null-space projection repairs suspect LoRA adapters, while expert quarantine changes training so triggered behaviour can be switched off. Results span constructed attacks and open 7B–13B models, not production supply chains.
Technology · Latest source 2026-11-05
Researchers saw answer-pattern differences, unexpected tool calls and five services reporting thousands more input tokens than the official APIs for identical requests. But the non-random, two-round audit could not see which models or wrappers ran behind each relay, and its price-group comparison did not reach conventional statistical significance.
Society & Media · Latest source 2026-11-07
A same-day Chinese report assembles newsroom examples and surveys from around the world, arguing that AI is becoming media infrastructure rather than a stand-alone tool. Its practical warnings about skills, verification and unequal access matter, but its case count, provider claims and in-house efficiency figures cannot be treated as an independent measurement of what works.
Government & Policy · Latest source 2026-11-05
The White House task force still lacks a published charter, but the Justice Department has now released a related internal memorandum covering investigations, charges and enforcement guidance. It coordinates DOJ work without creating a new offence, enforcement power or enforceable right.
Science & Research · Latest source 2026-11-04
An unreviewed three-country project used Claude agents to build a 35,562-line Lean library for inverse-problem theory. Lean checked every proof, while deterministic tests exposed one textbook theorem whose hypothesis was far narrower than it appeared—but fewer than half the stated theorems had direct author review.
Technology · Latest source 2026-11-05
Google made its video-generation and editing model generally available at $0.07 per second for 720p output, 30% below the previous version, and extended native clips from ten to 15 seconds. The release is broad and practical, but its quality claims rely on selected demonstrations and customer quotes rather than a published comparative benchmark.
Security & Defence · Latest source 2026-11-06
CERT-In's 107-page framework grades AI systems by impact, autonomy, data, scale and other risks, then links each tier to stronger audit expectations. It is the agency's most detailed AI-audit playbook yet, but the document does not identify a legal basis, start date or penalty for non-compliance.
AI Risks & Safety · Latest source 2026-11-06
Four-agent teams coordinated complex missions in EVE Frontier, but a prompted saboteur usually blended mistakes with plausible excuses. Separate message-only monitors detected 20–40% at a 5% false-accusation setting, while simulated traders exposed different weaknesses in risk-taking and market prediction.
Science & Research · Latest source 2026-11-06
OpenAI's GPT-6 Astra earned 66.7% of available credit at xhigh effort across gels, microscopy, sequencing and other raw evidence, versus 43.8% for GPT-5.6 Sol. The company benchmark shows progress, but its own examples still include wrong controls, reversed interpretations and inaccurate colony counts.
Technology · Latest source 2026-11-06
Persistent notes improved one 30-question benchmark, while a wiki built over 128 sessions helped with some multi-document questions. But stored knowledge could become stale, overwrite itself or grow larger than the source material, making selective memory and verification as important as recall.
Education · Latest source 2026-11-06
A national inquiry says short, unsupervised take-home exams are increasingly unreliable when generative AI is readily available. It recommends more controlled written and oral assessment, wider AI competence and strict conditions before automating grade explanations, but the proposals are advisory and now enter consultation.
Health & Life Sciences · Latest source 2026-11-06
Same-day note closure was 10.2 percentage points higher with an AI scribe in non-English language-concordant visits and 1.7 points higher in English visits after adjustment. Interpreter-mediated estimates were too imprecise for a conclusion, and the study did not test note accuracy or patient outcomes.
Security & Defence · Latest source 2026-11-06
DeepMind agents helped reproduce a restricted signature attack in 391 CPU core-years and derive a cheaper theoretical method for factoring many same-size keys. Both papers are working drafts, and neither demonstrates a break of today's mainstream RSA deployments.
Work & Skills · Latest source 2026-11-05
An audit of Terminal-Bench 2.0 found instruction mismatches, dependency drift and resource failures. The repaired 2.1 release kept the broad ranking but shifted individual scores by as much as 12 points and positions by three places.
AI Risks & Safety · Latest source 2026-11-05
VLA-Tug read a robot policy's internal states and often raised earlier, more informative alarms. Its guarantees held on exchangeable seen episodes; under unseen-task shift, target rates failed in some runs and steering remained heuristic.
Science & Research · Latest source 2026-11-05
Scientist World converted 1,000 papers into executable tasks, but its headline comparison ran frontier agents on a fixed ten-environment cohort. A positive score meant beating the reproduced method on hidden metrics—not validating a new scientific discovery.
Work & Skills · Latest source 2026-11-05
Seven coding agents tried to convert 20 PyTorch repositories to JAX on TPUs. The best passed 15; every other agent passed nine or fewer, and list-price cost ranged from $10 to $1,844 across each agent's 20 attempts.
Health & Life Sciences · Latest source 2026-11-05
A model trained on MIMIC-IV generated alternative ICU trajectories and reproduced one 24-hour recovery contrast missed by compact state models. Staffing, arrivals and service times were scenario assumptions, not measured hospital workflows.
Science & Research · Latest source 2026-11-05
SwarmEvolve let many agents exchange internal model states instead of prose and led the average score with three Qwen3 sizes. The study used only three seeds per problem, automatic evaluators and compatible open-weight models.
Health & Life Sciences · Latest source 2026-11-05
A Chinese preprint trained on 339,563 paediatric ECGs and reached a mean AUC of 0.9695 after fine-tuning. The improvement over paediatric training from scratch was only 0.0033, and the clinician test used 100 selected cases from the same hospital.
Climate & Energy · Latest source 2026-11-05
GenToka improved average error over a prior model and raised disruption-window F1 by 8.4% on one UK tokamak's historical data. It has not controlled a plasma, transferred to another machine or advanced net-energy fusion.
Technology · Latest source 2026-11-04
An unreviewed study marked about 18% of audited output as removable and pruned 41–46% of reasoning in two online tests. One benchmark lost 2.7 percentage points of task success, and an LLM judge defines what counts as waste.
Society & Media · Latest source 2026-11-06
The Commission says TikTok did not adequately assess or mitigate repeated recommendations involving mental health and body image. The finding is preliminary: TikTok disputes it, no fine has been imposed and the release does not measure harm in a new user cohort.
Technology · Latest source 2026-11-03
In 126,473 synthetic retrieval trials, neutral filler often cost almost nothing while rival statements sharply reduced accuracy. The study is an unreviewed preprint on small open models, not proof about production assistants.
Education · Latest source 2026-11-05
The Commission proposes AI-literacy assessment, shared education infrastructure and stronger teacher preparation. The package is a policy programme, not evidence that learning or teacher shortages have improved.