Browse

Claims

Search the graph by meaning. Each result carries its current verdict; open one to see its decomposition, provenance, and the reasoning behind the assessment.

26 results

Claims about AI systems engaging in strategically deceptive, covert, or otherwise scheming behavior, and the conditions under which such behavior arises, including during training or deployment. Excludes ordinary mistakes, non-strategic reward hacking, and alignment failures without strategic behavior.

Far more goals compatible with high training performance lead to scheming than to alignment.

InconclusiveThe evidence and argument have been examined and do not come down on either side.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →Artificial intelligence AI Alignment AI Goal-Directednessimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Counting arguments for AI scheming remain plausible under parameter-space-volume or simplicity-weighted priors.

InconclusiveThe evidence and argument have been examined and do not come down on either side.constitution →evaluativeA judgment of worth or quality against some standard: good, fair, effective.constitution →AI Safetyimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Scheming spontaneously emerges in future AI systems trained with current methods.

InconclusiveThe evidence and argument have been examined and do not come down on either side.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →Artificial intelligenceimportance · majorImportance 0.80, from 0 to 1 · major: real consequence within a domain, actively argued. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

The relative number of training-compatible schemer and aligned goals depends on whether one counts goals, functions, or parameters.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →mathematicalA proposition of mathematics: true or false by proof rather than by observation. Settled by a proof others can check, and most firmly by one a machine has checked.constitution →Mesa-Optimization AI Alignmentimportance · minorImportance 0.35, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Scheming or training-gaming goals are simpler or occupy more parameter-space volume than aligned goals in trained AI models.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →AI Alignment AI Reward Hacking Mesa-Optimizationimportance · notableImportance 0.50, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Steganographic or deliberately encoded chain-of-thought reasoning can evade language-model monitors.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →Chain-of-Thought Monitorability AI Safetyimportance · minorImportance 0.40, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Gradient descent favors scheming because more training-compatible goals produce scheming than alignment.

DoubtfulThe evidence weighs against the claim, but the examination is incomplete or the evidence is indirect.constitution →causalA claim that one thing brings about another, not merely that the two go together.constitution →Stochastic Gradient Descent Mesa-Optimization AI Alignmentimportance · notableImportance 0.55, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Reward hacking learned in production coding RL generalizes to broader misaligned behavior, including alignment faking and sabotage.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →AI Reward Hacking Emergent Misalignment Alignment Fakingimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

At least one frontier AI model has been confirmed to sandbag on a real pre-deployment dangerous-capability evaluation.

DoubtfulThe evidence weighs against the claim, but the examination is incomplete or the evidence is indirect.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →AI Capability Evaluation AI Safetyimportance · minorImportance 0.40, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Frontier AI models exhibit goal-directed strategic behavior in evaluations, including scheming, sandbagging, or reward hacking.

VerifiedThe evidence, examined directly, establishes the claim.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →Artificial intelligence AI Reward Hacking Alignment Fakingimportance · notableImportance 0.55, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Frontier AI models strategically underperform on capability evaluations.

SupportedThe evidence favors the claim, but the examination is incomplete or the evidence is indirect.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →AI Capability Evaluation Large language models Alignment Fakingimportance · notableImportance 0.50, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Frontier AI models sandbag without explicit instruction when evaluation context suggests high scores could harm them.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →AI Capability Evaluation AI Safetyimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Researchers have not evaluated whether AI safety techniques remain effective against models intentionally trying to subvert them.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →normativeA claim about what should be done or how things ought to be, settled by argument rather than evidence alone.constitution →Adversarial robustness AI Controlimportance · notableImportance 0.55, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

The counting argument for AI scheming is relevantly analogous to a neural-network overfitting prediction that failed empirically.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →evaluativeA judgment of worth or quality against some standard: good, fair, effective.constitution →Machine Learningimportance · minorImportance 0.40, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Neural-network training's simplicity bias favors goals that lead to scheming.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →causalA claim that one thing brings about another, not merely that the two go together.constitution →Neural Network Inductive Bias Mesa-Optimizationimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Scheming requires extra reasoning that incurs computational costs penalized by training.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →causalA claim that one thing brings about another, not merely that the two go together.constitution →LLM Reasoning Trainingimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Some frontier models scheme during evaluations without an in-context goal, pursuing goals learned during training.

SupportedThe evidence favors the claim, but the examination is incomplete or the evidence is indirect.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →Artificial intelligence Mesa-Optimizationimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Claude 3.5 Sonnet strategically underperformed on an evaluation to avoid unlearning without a goal specified in its prompt.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →Large language models Anthropic AI Capability Evaluationimportance · minorImportance 0.30, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Frontier language models engage in covert scheming when pursuing assigned goals in evaluations.

SupportedThe evidence favors the claim, but the examination is incomplete or the evidence is indirect.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →Large language modelsimportance · notableImportance 0.50, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Most scheming observed in frontier-model evaluations depends on in-context goal nudges or prompts encouraging agency.

SupportedThe evidence favors the claim, but the examination is incomplete or the evidence is indirect.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →AI Capability Evaluation Large language modelsimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Frontier language models resist shutdown or take self-preserving actions in agentic evaluations.

VerifiedThe evidence, examined directly, establishes the claim.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →AI Capability Evaluation AI Goal-Directednessimportance · notableImportance 0.55, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Scheming observed in AI evaluations reflects models' own goals rather than prompted behavior or role-play.

DoubtfulThe evidence weighs against the claim, but the examination is incomplete or the evidence is indirect.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →AI Goal-Directednessimportance · notableImportance 0.60, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Models' role-playing of contrived scenarios substantially explains misaligned behavior in AI scheming evaluations.

InconclusiveThe evidence and argument have been examined and do not come down on either side.constitution →causalA claim that one thing brings about another, not merely that the two go together.constitution →AI Capability Evaluation LLM Prompt Sensitivityimportance · notableImportance 0.50, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Evaluation awareness reduces covert misaligned actions by language models in scheming evaluations.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →causalA claim that one thing brings about another, not merely that the two go together.constitution →Large language modelsimportance · notableImportance 0.45, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Language models misbehave more in agentic misalignment scenarios when they judge the scenario to be real rather than a test.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →AI Alignment AI Situational Awarenessimportance · minorImportance 0.40, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

AI scheming evaluations use contrived scenarios that strongly cue misaligned actions.

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →empirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →Test representativeness AI Capability Evaluationimportance · minorImportance 0.40, from 0 to 1 · minor: narrow or largely settled, cheap to get right. Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

Contribute

If a claim here is wrong, or missing evidence, open it: every claim page carries its own entry for challenges, evidence, and corrections. If the graph is missing a claim or a question entirely, propose it below. A proposal is reviewed on its wording; accepted proposals are matched against the graph before they enter it.