Everyone Has Updated Their AGI Timeline. Nobody Has Updated Their Plans.

In 2020, the consensus among AI researchers on when transformative AI might arrive was something like "probably not this century, maybe not ever in the way people imagine." Survey after survey of ML researchers put the median estimate for human-level AI at 2050 or later. Forecasting platforms like Metaculus had community medians in the 2040s. The people doing the work didn't think it was imminent.

In 2025, the same surveys and platforms showed median estimates in the range of 2028–2032. The expected timeline collapsed by about two decades in five years.

The safety budget has not moved proportionally. The money going into alignment research, governance development, and deployment oversight is roughly where it was.

What the Forecasts Say

Metaculus, the community forecasting platform with one of the better track records on technological questions, tracks several AGI-related questions. As of early 2026, the median resolution date for "weak AGI" (a system that can match human performance on almost all cognitive tasks) is around 2030. For "transformative AI" (AI that causes GDP growth comparable to the industrial revolution), the median is in the early 2030s.

The Alignment Forum's annual survey of AI safety researchers, the people who think most carefully about this for a living, shows similar compression. The median of expert estimates for when the first AI system capable of performing at human level across most cognitive tasks will exist moved from 2042 (in 2021) to about 2030 (in 2025). A decade of expected runway, gone in four years.

These aren't wild extrapolations from hype. They are forecasts from people tracking the rate of capability improvement and updating their models accordingly. The most detailed attempt to turn those forecasts into a concrete, dated sequence of events is the AI 2027 scenario, which names the specific capabilities that would have to develop, and roughly when.

Why Timelines Compressed

The capabilities that drove the update were real and documentable. GPT-4 in 2023 passed the bar exam, the USMLE (the US medical licensing exam), and a range of graduate-level professional assessments that had previously been assumed to require years of domain expertise. Claude 3 Opus, released in early 2024, showed performance improvements across reasoning tasks that surprised even people at the labs that built it. The GPT-5 and Claude 4 releases in 2025 continued the trend.

The scaling laws held. With more compute, more data, and better architectures, capability improvements continued at rates that exceeded many models of how fast they could continue. People updated their forecasts because the evidence demanded it.

The Money Tells the Story

Global AI safety research funding, counting Anthropic, DeepMind's safety team, OpenAI's superalignment efforts (which were significantly reduced in 2024), MIRI, ARC, and a collection of academics, is somewhere around $500 million per year. That sounds substantial until you compare it to AI capability investment, which runs to tens of billions of dollars annually and is growing.

The ratio of capability investment to safety investment is approximately 100:1. That ratio has been getting worse, not better, as capability investment has accelerated and safety teams at major labs have faced layoffs and restructuring.

The governance situation is similarly behind the curve. The EU AI Act, the most comprehensive regulatory framework to date, was primarily designed for AI systems in the capability range of 2022. Frontier systems are already beyond several of its assumptions. The US executive orders on AI safety have been partially implemented, partially reversed, and don't have independent technical expertise to enforce them against frontier capability claims.

Sam Altman Said the Quiet Part Out Loud

In a 2024 blog post, Sam Altman wrote about "superintelligence in a few thousand days," which is roughly a decade. Here was the CEO of the most prominent AI lab stating publicly that transformative AI is a near-term planning horizon rather than a speculative future. He framed it optimistically, in terms of abundance, medical breakthroughs, and economic growth.

What the framing omitted is that a near-term planning horizon for capability has to be a near-term planning horizon for safety. You can't treat "we'll deploy transformative AI in a decade" as opportunity while treating "we need to ensure that AI is aligned before deployment" as a research project that can continue at its current pace.

The math is simple enough to do on a napkin. If AGI arrives in 2030, the alignment problem requires twenty years of research to solve, and it's currently 2026, the timeline is already wrong. The solution could never arrive in time. Alignment may well be solvable; the problem is that the available time is being spent elsewhere, and an intelligence explosion could compress what time remains far faster than the calendar suggests.

Plans That Would Match the Forecasts

Safety investment would scale with capability investment. Governance would develop at the pace of capability. Deployment decisions would turn on safety evaluations rather than competitive timelines that no one can unilaterally exit.

None of that describes the current situation. The timelines have moved; the plans have not. Maybe this is manageable. Maybe alignment is easier than we fear, or the timeline forecasts are still off. Or this is the most important planning failure in human history. The stakes are high enough that "maybe it's fine" is an unreasonably casual attitude to take into a decade like this one.