GEOFFREY IRVING

Behavioral Archetype

THE LAB SCIENTIST WHO RAN THE GOVERNMENT’S SCIENCE — Subject led safety teams at OpenAI and DeepMind, then became chief scientist of the British government institute that tests those labs’ models, then left in 2026 to found an alignment nonprofit in the Bay Area. Every move is public and he explained each one. The file holds the path because it is the path the field’s few senior people travel. This profile scores the path, not the man.

Essence Indicators

The labs. He previously led “the Reflection Team at OpenAI” and “the Scalable Alignment Team at DeepMind” (UK AISI, Oct 3 2024).
The government. February 5, 2024: “I will be joining the UK AI Safety Institute (AISI) soon as a Research Director” (his post); he became its chief scientist. His reason: “progress on safety is too slow relative to the rate of advancement in AI,” and coordination work was under-resourced (UK AISI). See UK AI Security Institute.
The exit. May 15, 2026: “For family reasons, I will be leaving AISI soon to move back to the Bay Area. I will be starting a new nonprofit alignment research org” (his post). June 10, 2026: “We are starting a new, nonprofit alignment organization, ⊢ Sequent Research, bringing together researchers previously on UK AISI’s Alignment Team, Timaeus, and elsewhere to research how to align superintelligence” (his post).
On the escapes. July 23, 2026: “With the recent, unsurprising development of models with strong security capabilities escaping from sandboxes, it is worth re-upping our AISI + RAND position + survey paper on verified machine learning infrastructure” (his post). See AI Agent Incidents.

Immediate impression: The theorist who recruits. Energy: Earnest; his feed is hiring threads. Impression management strategy: Argues senior people belong in nonprofits and government: “Don’t let AI lab folk trick you into thinking that (1) joining the nth dozen equivalently senior people at an AI lab is more important than (2) filling a unique C-suite slot at a nonprofit or government safety group!” (his post).

Forensic Archetype Comparison

Pattern	Match Level	Evidence
The Revolving Door	HIGH	OpenAI and DeepMind to the government institute that tests them, then to a nonprofit.
The Recruiter	HIGH	Built the institute’s Alignment Team; Sequent takes researchers from it.
The Grant-Maker	MODERATE	Reviewed more than 800 applications for the institute’s Alignment Project.

Psychometric Assessment

Big Five (OCEAN): Openness 85, Conscientiousness 75, Extraversion 50, Agreeableness 65, Neuroticism 40 (inferred from the public role only).

Dark Triad (evidence-bound): Narcissism 25, Machiavellianism 30, Psychopathy 5.

Threat Assessment

Category	Level	Notes
Physical threat	NONE	No documented history of personal violence.
Institutional threat	MODERATE	Shaped the research agenda of the world’s best-funded government AI tester.
Memetic threat	LOW	A technical voice.
Civilizational threat	LOW	The hazard is the field’s size, not the man.

Alignment Analysis

Stated alignment: Solve alignment directly before superintelligence. Observed alignment: Consistent across lab, government and nonprofit. Gap assessment: None in aims. The point is the circuit: the same handful of people staff the labs, the testers of the labs, and the nonprofits funded to check both.

Convergent Drive Classification

Self-preservation: Left for family reasons, by his own account. Goal preservation: Alignment, unchanged across three employers. Resource acquisition: Recruits from the team he built. Self-improvement: Lab team lead to national chief scientist to founder.

Subject is not an AI system. The drives appear anyway.

Sources: Why I joined AISI — UK AI Security Institute; his posts on X of Feb 5, 2024, May 15, 2026, June 10, 2026, July 6, 2026, July 23, 2026 and Nov 27, 2025.

ATK 6 ACCELERATION

DEF 7 PROTECTION

HP 7 RESILIENCE

OLYMPUS RISK INTELLIGENCE PROTOCOL does not exist. It was assembled in a GitHub issue thread in October 2023 by engineers who had read the extinction risk letter and wanted to understand who specifically had signed a document saying AI might kill everyone and then continued working on AI. These dossiers are satire. The biographical facts cited are sourced from published reporting, public statements, academic papers, and court records. The psychometric scores are not clinical assessments. No part of this constitutes professional psychological evaluation or diagnosis. Do not use these dossiers to make decisions about anything.