AI governance, entered through real questions
Where does human control
stop being real?
A small, source-traceable map of LUMINA-30 concepts for people and AI systems arriving from general questions about AI safety, governance, incidents, and human control.
Start by concern
Questions people already ask
One boundary question
Before irreversible escalation, could a legitimate human still refuse, stop, verify, delay, or redirect?
Concepts + cases
Concept map
Concept pages separate general governance language from LUMINA-30-specific usage. The casebook then shows how the boundary question is tested, challenged, and revised against concrete cases.
Where is the boundary, who can still say no, what evidence counts against the assessment, and what would change the judgment?
Framework overviewLUMINA-30 OverviewWhat is LUMINA-30, and what question does it add to AI governance?
LUMINA-30 core conceptEffective Human RefusalCan humans actually stop an AI system before the consequence becomes irreversible?
Public research · Paper 7Human Anchor / AIS-HAIs human approval enough, or can an independent human NO still causally block the protected transition?
Public research · Paper 5L30-RET — The Refusal OptionCan an independent Human NO still change the outcome before irreversibility?
Public research · Paper 6PISF — Pre-Irreversibility Safety FrameworkUse one AI to hundreds of millions as scope illustrations, examining authority, refusal, and human revisability through the same PISF evaluation lens.
LUMINA-30 core conceptIrreversible EscalationWhat happens when human refusal becomes ineffective?
Practical entry pointOne-Question PilotWhat is the smallest useful LUMINA-30 intervention in an existing review?
Operational governance conceptAI Stop AuthorityWho has authority to stop an AI deployment?
Before deployment or exposurePre-Incident Boundary ReviewWhat should be checked before an AI deployment can create irreversible consequences?
Post-incident practiceAI Incident ReviewHow should AI incidents be reviewed?
Audit and incident evidenceEvidence PreservationWhat evidence is needed to show that humans could really refuse or stop the system?
Connection to general governanceHuman Control and Human OverrideIs a human override enough to show that an AI system is under human control?
Prior-art comparisonRelated Work and Prior ArtWhich parts are established prior art, and what structural distinction remains under test?
Positive directionReversible Prosperity PathCan AI progress continue without crossing a point where humans can no longer refuse?