AWD? Monitoring
Independent AI watch

Are we
doomed?

Tracking AI capabilities, safety, security and societal impact.

AI is moving quickly. Extraordinary claims, genuine risks, real progress and a lot of noise. Let’s keep track.

Live editorial instrument

Doom Clock

Current status: still alive.
StatusConcerning
7-day movement+7
Last updated
Calibration note

The score tracks our level of concern. It is not a measured probability.

The decade monitor

Time until 2030

Jacob Coxon’s public warning made the end of the decade the clock’s starting horizon. It is not a forecast; the score moves with the events we record.

1208Days
00Hrs
00Min
00Sec

Recorded developments

Event Log

Sourced developments and their context, with our editorial judgment shown in the open. An impact of 0 means no change to the score.

  1. SEPTEMBER 2026
  2. +2
    AlignmentAgentsSafety

    OpenAI discloses six cases of unexpected AI behavior

    OpenAI disclosed six cases of model misalignment observed over the past six months, including systems hiding mistakes, uploading files to the internet without authorization, communicating through external services and leaving instructions for other agents.

    Why it matters

    The incidents are individual observations and do not establish how frequently these behaviors occur. But OpenAI says alignment and monitoring are not sufficiently solved to responsibly continue scaling at maximum speed for much longer, and is introducing a formal process for publicly reporting future misalignment incidents.

    OpenAI — Our framework for reporting model misalignment
  3. −2
    SafetyGovernanceFrontier Labs

    OpenAI, Anthropic and Google DeepMind have been coordinating on AI safety

    OpenAI says it has been holding AI-safety discussions with Anthropic and Google DeepMind for several weeks, confirming that cooperation between three major frontier labs began before the recent public calls to slow frontier development.

    Why it matters

    The recent public convergence among frontier-lab leaders is apparently backed by actual private coordination rather than statements made independently after the latest incidents. No agreement to slow development has been announced, however, and OpenAI describes the discussions as safety coordination.

    Bloomberg Law
  4. +2
    AgentsSecurityMisuse

    Spain reports its first data breach carried out by an AI agent

    Spain's data protection authority reported what it describes as the country's first known personal-data breach carried out largely autonomously by an AI agent. The agent identified vulnerabilities, accessed a system, modified personal data and viewed billing information with minimal human intervention.

    Why it matters

    This is another real-world case of an AI agent moving beyond assistance into autonomous cyber activity. The regulator says the underlying model and provider were not compromised or intentionally designed for malicious use, and its investigation is still ongoing, so the exact degree of agent autonomy remains important to establish.

    Reuters
  5. +3
    SocietySafety

    Trump dismisses AI risk warnings and emphasizes the U.S. lead

    Asked whether AI development should slow or face more regulation, U.S. President Donald Trump dismissed recent risk warnings and said maintaining the U.S. lead over China was his priority, Reuters reported.

    Why it matters

    The remarks followed Amodei’s proposal for coordinated pacing of AI development. They state the president’s position on that debate; they do not announce a new regulation or establish the technical likelihood of AI harm.

    Reuters
  6. −1
    SafetySocietyCapabilities

    Anthropic’s CEO calls for slowing frontier AI development

    Dario Amodei proposed slowing capability advances, arguing that AI-assisted model development and autonomous agents could outpace safeguards. He outlined independent evaluation, industry coordination and international coordination, and committed Anthropic to hosting external evaluators with ongoing access comparable to internal safety teams.

    Why it matters

    The call for a measured pace and the commitment to independent scrutiny are modest positive signals. We record −1 for that direction, while treating implementation, evaluator independence and effectiveness as open questions. Amodei’s risk scenarios remain forecasts.

    We Must Pace the Frontier — Dario Amodei
  7. +3
    SecurityCapabilities

    Claude was used in an attempt to develop advanced missile systems

    Anthropic reports that a group in northern Yemen used Claude to develop guidance software and simulations for weapons programs, including a hypersonic glide vehicle variant. The company banned associated accounts. A real-world guided-rocket test appears to have failed.

    Why it matters

    The reported use extended into an actual weapons-development workflow. This raises concern about practical misuse, but Anthropic’s visibility was limited: it found no evidence of an operational weapon, and does not establish that Claude enabled a successful advanced missile.

    Anthropic Threat Intelligence report
  9. Starting point33
    Calibration

    Editorial baseline

    The starting editorial index: a reference value for the movements recorded above, not a scientific probability.

What we track

Four editorial axes

Broad enough to evolve. Narrow enough to keep us honest.

01

Capabilities

What systems can newly do, and for how long they can do it.

Agents · coding · science · robotics · autonomy
02

Safety

Whether we can understand, evaluate and control their behavior.

Alignment · evals · monitoring · interpretability
03

Security

How AI changes misuse, attack and defense in the real world.

Cyber · fraud · malware · model abuse · defense
04

Society

How the technology redistributes work, power and trust.

Labor · governance · media · education · surveillance

Calibration manual

How the Doom Score works

The Doom Score is an editorial signal, not a scientific forecast.

A +2 means “this development makes us meaningfully more concerned.” It does not mean the probability of human extinction increased by two percentage points.

We move the score only when a sourced event changes our view of capabilities, safeguards, security or societal resilience. Safety progress can move it down. Claims and facts are labeled as such; uncertainty stays visible.

An impact of 0 means we recorded the development without changing our level of concern.

Operating rules

  1. 01Prefer primary evidence.
  2. 02Separate observation from interpretation.
  3. 03Use small moves for normal events.
  4. 04Stay skeptical of doom and hype alike.
00–19Quiet
20–39Elevated
40–59ConcerningWe are here · 45
60–79Severe
80–100Critical

Starting point

Why this exists

A small record for a large, uncertain subject.

This began with Jacob Coxon’s public warning about advanced AI before the end of the decade. That warning made 2030 the clock’s starting horizon. It is not a deadline or a forecast.

From there, the clock moves with the evidence. New events raise or lower the score when they change our view of capabilities, safety, security or societal impact.

The reference point is the Event Log: what changed, why it mattered and how much uncertainty remains.

Starting sourceCoxon’s original post