The archive
1,158 AI reports, with their evidence attached
Every report published by The Impact of AI between June and November 2026 — each one carrying its primary source, its stated limits and what it means for people.
Latest
All 1,158 reportsCan a cheaper AI API relay change what you get—or what you pay for? A 20-service audit found warning signs, not proof of fraud
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.
Can AI agents read routine wet-lab evidence? A 100-task benchmark left Astra short of full credit
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.
Climate & EnergyCan AI forecast a city’s power demand from ten annual observations? Tiny reported errors rest on a two-year test
A peer-reviewed study forecasts electricity demand in Adigrat, Ethiopia, through 2035 with a PSO-optimised neural model. Its total-load error was reported as 0.027%, but the model learned from eight annual observations and was tested on only two.
Science & ResearchCan AI formalize mathematics without humans reading every proof? AutoIP shifts the review burden to statements
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.
Health & Life SciencesCan AI models answer infectious-disease pharmacy exams? Five scored near 90%—but images and harder reasoning exposed gaps
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.
Recent reports
Archive- Can AI-agent teams detect a saboteur? In Anthropic's game tests, almost neverAI Risks & Safety
- Can an AI agent call every app on a phone? China’s new network-access notice puts consent and compatibility firstTechnology
- Can thermodynamics show what an AI learns about event order? A new measure found a signal—but not causal understandingScience & Research
- Can two defences stop hidden backdoors in fine-tuned AI? Lab attacks fell sharply—but real compromised models were not testedSecurity & Defence
- Did AI break RSA? It accelerated two attacks on legacy 1024-bit settings—not modern 2048-bit encryptionSecurity & Defence
- Do AI agents amplify social contagion? Five models adopted more readily—but the spread was simulatedSociety & Media
- Do AI scribes help equally across patient languages? A 209,009-encounter study found uneven associationsHealth & Life Sciences
- Do India’s new AI audit guidelines create binding rules? They set a four-tier framework—but legal force is unclearSecurity & Defence
- Does AI help manufacturers make greener products? A 333-firm survey finds an association—not emissions proofFinance & Business
- Does more memory make AI agents better? Anthropic's tests found useful gains—and costly overgrowthTechnology
- What changed in Google’s Gemini Omni 1.2 video model? Longer clips and lower prices arrived without an independent quality testTechnology
- What will Japan's emergency cyber package change—and where does AI fit?Security & Defence
Browse by section
The desk
About the desk
Who writes these reports, and the three questions every one answers.
Editorial standards
How sources are verified, how limits are stated, and how the impact record is scored.
Impact tracker
Every record, by evidence stage — announced, funded, deployed or reviewed.
The Impact Brief
Three findings and one evidence visual, prepared weekly.
Guides
Longer readouts for education, healthcare and research tools.
Watch
Video and short form on the same reporting standards.