How DoomBench scores the news
DoomBench is a structured editorial index. It is designed to make qualitative judgments visible and comparable, not to disguise them as scientific probabilities.
What the index means
Zero represents a world with no meaningful AI capability. One hundred represents the terminal AI takeover scenario associated with loss of human control. Neither endpoint is a probability claim.
The chronology begins at an editorial baseline of 12 on 1 January 2020. That baseline acknowledges that AI already existed before the tracked news window without treating earlier history as zero.
How one item moves the score
Direction is toward or away from doom. Magnitude and evidential confidence are each scored from 0 to 100. The maximum movement from one item is 6 index points, preventing a single headline from overwhelming the longer evidence trail.
Company attribution
Every company materially involved in a story is recorded. When several companies are involved, the assessed effect is divided equally so the company comparison sums to the original news effect. Government, academic, and independent items remain in the index without being falsely classified as companies.
Version-specific model scores
Each named model version receives a separate 0 to 100 risk profile. Capability measures the tasks it can perform. Autonomy measures sustained action with tools. Deployment measures practical access and diffusion. Misuse measures dangerous-use potential. Control difficulty measures residual risk after documented safeguards and access restrictions.
Model scores use the same endpoint meaning as the index, but they do not directly change it. A model launch affects the main Doom Index only through its separately sourced news assessment, preventing the same evidence from being counted twice.
Category taxonomy
Capability gains
Meaningful improvements in reasoning, science, coding, or general problem solving.
Autonomy and agency
Systems acting over longer horizons, using tools, computers, or physical machines.
Deployment reach
Access, affordability, adoption, and integration into consequential workflows.
Human displacement
Evidence that AI substitutes for, changes, or fails to replace human work.
Misuse and incidents
Real-world misuse, loss of control, cyber incidents, deception, or dangerous access.
Safety and alignment
Evaluations, safeguards, alignment work, and evidence about their limits.
Governance and control
Binding rules, oversight institutions, standards, and enforceable controls.
Competitive race
Pressures that accelerate development, diffusion, concentration, or strategic competition.
Human resilience
Evidence that people, institutions, or technical controls retain meaningful advantage.
Research and revision rules
- Prefer primary sources, then authoritative synthesis and credible reporting.
- Require an explicit, non-future publication date verified in Europe/Dublin.
- Consolidate syndication and substantially identical stories.
- Include counterevidence and safety progress, not only alarming developments.
- Recalculate chronologically when backfilled evidence changes the record.
- Append score revisions instead of silently overwriting earlier judgments.
- Preserve all durable records when discovery is partial, interrupted, or uncertain.