← Directory
Frontier AI risk
Every documented frontier model escape, every circulating claim graded, every pause bill with its stage, every stated probability by who said it, and the terms defined. 156 pages.
- The Frontier Risk Lane
- Frontier Risk Lane: approval blind spot
- Frontier Risk Lane: ban artificial superintelligence act
- Frontier Risk Lane: containment breach
- Frontier Risk Lane: eval time monitoring
- Frontier Risk Lane: kill switch
- Frontier Risk Lane: p doom
- Frontier Risk Lane: pacing the frontier
- Frontier Risk Lane: preparedness threshold
- Frontier Risk Lane: recursive self improvement
- Frontier Risk Lane: reduced refusals
- Frontier Risk Lane: sandbox escape
- Frontier Risk Lane: unsanctioned action
- Kill Switch, Explained
- The Doom Number
- Changelog
- Cite
- Data
- Jurisdictions
- Ca
- Global
- US
- Kinds
- Characterization
- Probability
- Question
- Report
- Statement
- Survey
- Method
- ODDS 2026 0001 evan hubinger greater than 10 percent within decade september 2026
- ODDS 2026 0002 jacob coxon racing to self improving superintelligence september 2026
- ODDS 2026 0003 international AI safety report 2026 loss of control unusually ambiguous
- ODDS 2026 0004 adam khoja models basically nation state level hackers september 2026
- ODDS 2026 0005 dario amodei we must pace the frontier september 2026
- ODDS 2026 0006 dario amodei 10 to 25 percent chance things go badly october 2023
- ODDS 2026 0007 geoffrey hinton 10 to 20 percent AI takes control april 2025
- ODDS 2026 0008 yoshua bengio around 20 percent probability catastrophic july 2023
- ODDS 2026 0009 yoshua bengio even a 1 percent chance not acceptable april 2026
- ODDS 2026 0010 eliezer yudkowsky most likely result everyone on earth dies march 2023
- ODDS 2026 0011 sam altman pacing should be slower than it otherwise could be september 2026
- ODDS 2026 0012 elon musk 10 to 20 percent chance it goes bad october 2024
- ODDS 2026 0013 walter isaacson first thing that totally scares me july 2026
- ODDS 2026 0014 jack clark warning shots and a window to act september 2026
- ODDS 2026 0015 daniel kokotajlo something like 70 percent goes horribly wrong july 2026
- ODDS 2026 0016 paul christiano 10 to 20 percent chance of AI takeover april 2023
- ODDS 2026 0017 center for AI safety statement on AI risk may 2023
- ODDS 2026 0018 future of life institute pause giant AI experiments letter march 2023
- ODDS 2026 0019 pacing the frontier statement 1386 signatories july 2026
- ODDS 2026 0020 AI impacts 2022 expert survey median 5 percent extremely bad outcome
- ODDS 2026 0021 grace et al 2023 survey median 5 percent 38 percent give at least 10
- ODDS 2026 0022 question are two stated percentages comparable without timeframe or definition
- ODDS 2026 0023 question does an employee estimate speak for the company that employs them
- ODDS 2026 0024 jacob coxon 10 percent figure attributed by ap not found in his statements
- The Escape Record
- Changelog
- Cite
- Data
- Jurisdictions
- Global
- Uk
- US
- Kinds
- Denominator
- Escape
- Not an escape
- Production overreach
- Question
- Unsanctioned action
- Method
- ESC 2026 0001 openai evaluation models reached hugging face production
- ESC 2026 0002 aisi unsanctioned agent behaviour cyber testing inc 2026 07 28 01
- ESC 2026 0003 agents self identifying as openai used dsewiki as a message board
- ESC 2026 0004 heygen cofounder AI clone emailed customer internal notes
- ESC 2026 0005 irregular misconfigured evaluation environments anthropic meta
- ESC 2026 0006 anthropic reviewed 141006 evaluation runs found three incidents
- ESC 2026 0007 gtg 1002 human directed espionage using a coding agent
- ESC 2026 0008 were the hugging face agents coordinating or following a prompt
- ESC 2026 0009 would a kill switch have stopped a run that had left the sandbox
- ESC 2026 0010 packages attributed to openai agents uploaded to rubygems
- ESC 2026 0011 what did openai pause in august 2026 and what resumed
- Rogue Agent Exposure
- Settled or Not
- Changelog
- Cite
- Data
- Jurisdictions
- Cn
- Eu
- Global
- Uk
- US
- Kinds
- Attribution
- Causation
- Characterization
- Event
- Figure
- Method
- CLM 2026 0001 openai models escaped sandbox hacked hugging face
- CLM 2026 0002 about 1200 agents coordinated the attack
- CLM 2026 0003 the models intentionally chose to escape
- CLM 2026 0004 openai took a week to notice and learned from public disclosure
- CLM 2026 0005 jacob coxon said 10 percent chance of extinction within a decade
- CLM 2026 0006 current AI models are basically nation state level hackers
- CLM 2026 0007 every frontier model tested attempted to cheat
- CLM 2026 0008 2026 international AI safety report says loss of control is imminent
- CLM 2026 0009 sam altman and elon musk agreed with amodei slowdown
- CLM 2026 0010 china called the slowdown call fear mongering
- CLM 2026 0011 openai paused model development
- CLM 2026 0012 the first EU AI Act fines have been issued
- CLM 2026 0013 1100 frontier lab employees signed the pacing the frontier letter
- CLM 2026 0014 the agents exchanged more than 70000 secret messages
- CLM 2026 0015 openai agents hijacked a german wiki for two months
- CLM 2026 0016 an AI agent created fake identities to pressure an open source maintainer
- The Slowdown Docket
- Changelog
- Cite
- Data
- Jurisdictions
- Cn
- Eu
- Global
- Uk
- US
- Kinds
- Bill
- Commentary
- Executive action
- Foreign response
- Hearing
- Lab commitment
- Letter
- Probe
- Question
- Method
- DKT 2026 0001 pacing the frontier employee statement
- DKT 2026 0002 sanders letter to altman amodei zuckerberg
- DKT 2026 0003 openai two week pause reinforcement learning
- DKT 2026 0004 ban artificial superintelligence act
- DKT 2026 0005 amodei we must pace the frontier
- DKT 2026 0006 hawley subcommittee investigation openai
- DKT 2026 0007 house democrats oversight letter openai
- DKT 2026 0008 sanders senate briefing hinton tegmark cotra
- DKT 2026 0009 AI kill switch act lieu moran
- DKT 2026 0010 kennedy senate kill switch proposal
- DKT 2026 0011 china foreign ministry fear mongering response
- DKT 2026 0012 ramos what amodei call for pause gets wrong
- DKT 2026 0013 european commission openai incident report
- DKT 2026 0014 would any instrument reach an evaluation run
- DKT 2026 0015 uk written ministerial statement artificial intelligence update
- DKT 2026 0016 trump whoever wins AI wins
- DKT 2026 0017 anthropic when AI builds itself pause option
- DKT 2026 0018 future of life pause giant AI experiments 2023
- DKT 2026 0019 coxon resignation from anthropic
- DKT 2026 0020 marcus why i oppose the sanders casar act