THE TIMELINE

A widely repeated projection, quoted for years as though it were a measurement. It did not happen: by 2026 the machine-written share of new articles had plateaued near half. Kept at the top of this timeline deliberately.

The architecture underneath was five years old. What changed was a text box anyone could type into — which is why this date, not the research, is the starting line for nearly every measurement on this site.

Microsoft Research published VALL-E, producing convincing speech from a three-second recording of a speaker it had never heard, preserving emotion and the acoustics of the sample. The code was never released. The capability arrived anyway.

Stanford researchers tested seven detectors on 91 TOEFL essays and found an average false-positive rate of 61.2%, against near-perfect accuracy on native-written essays. Schools were already using these tools to accuse students.

A finance employee in Hong Kong authorised fifteen transfers to five accounts after a video meeting in which every other participant was synthetic. Police briefed publicly in February; the firm was identified in May and confirmed no internal systems had been breached.

Bots reached 51% of measured web traffic — the first crossing in a decade of monitoring.

The RAID benchmark evaluated twelve detectors against 6.2 million generations across eleven models, eight domains and eleven adversarial attacks. Detectors generalised poorly to unseen models, and simple changes — synonym swaps, a repetition penalty — produced steep accuracy drops. The paper opens by noting commercial detectors advertise 99%+ accuracy.

Analysis of English-language articles via public crawl data found AI-generated articles passing human-written ones in volume — then plateauing near half rather than continuing to climb.

Energy and AI put global datacentre electricity demand at roughly 945 TWh by 2030 — just under 3% of world electricity — against about 415 TWh in 2024. A large datacentre, the agency noted, can consume as much power as 100,000 households; the largest under construction, as much as 2 million.

California's SB 243 set disclosure requirements, break reminders for minors and self-harm protocols for companion apps, effective 1 January 2026.

The second consecutive year machines outnumbered people online. Human activity fell from 49% to 47%; malicious bots alone accounted for 40% of all traffic.

SB 942 begins requiring disclosure of AI-generated content, alongside SB 243's companion-chatbot rules.

A benchmark of 480 tasks drawn from investment banking, consulting and corporate law found the best model completed about 24% at first attempt, and roughly 40% after eight. Performance degraded sharply with task length.

xAI merged into SpaceX in an all-stock deal, and was dissolved as a separate company that May. The clearest case so far of frontier AI consolidating into heavy industry — which changes who answers for a model's behaviour.

In roughly twenty-five years of the Internet Crime Complaint Center's history, artificial intelligence appeared as a tracked category for the first time: 22,364 complaints and 93,346,472 in reported losses for 2025. Americans over 60 reported roughly .7 billion in losses across all internet crime, the steepest of any age group. The bureau notes the AI figure is almost certainly an undercount — it counts only cases where the victim recognised AI was involved.

University researchers demonstrated that every major agent benchmark could be driven to near-perfect scores without solving a single task, by attacking the evaluation harness rather than the problem.

The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force six days before the AI Act's main application date. It deferred the high-risk obligations to December 2027 and August 2028, and it added a prohibition on AI systems generating child sexual abuse material or non-consensual intimate content, plus expanded central enforcement powers for the AI Office. Coverage reported the delay; the additions received far less attention.

Article 50 of the EU AI Act begins requiring providers of generative systems to mark outputs in a machine-readable format, with penalties reaching millions of euros or a share of global turnover. The first broad legal requirement of its kind.