SAM ALTMAN
Behavioral Archetype
THE ACCELERATION ENGINE — Subject converts every constraint — nonprofit governance, safety culture, board oversight, regulatory attention — into additional growth capital and expanded deployment authority. The mechanism is not deception; it is transformation. He takes the thing that is supposed to slow him down and turns it into a thing that makes him go faster. This has worked every time.
Essence Indicators
- Survived a board coup in November 2023 — fired Monday, reinstated Friday, employees who engineered his return now running the company
- Co-authored non-disparagement agreements, publicly called them “genuinely embarrassing,” did not remove them until forced, continued as CEO
- Signed the 2023 extinction risk letter warning that AI poses a civilizational threat; continued building
- Publicly describes himself as “a bit of a believer” in AGI; has also said the company needs to generate revenue; appears to experience no cognitive dissonance from the juxtaposition
- Mission statement has been revised six times in nine years; “openly share,” “as a whole,” “unconstrained by financial return,” and “safely” have each been removed
Social Persona / Impression Management
Immediate impression: Calm, thoughtful, visionary. Projects intellectual engagement and earnest concern for humanity. Does not appear nervous in any documented public appearance.
Energy: Collaborative on the surface, extractive in aggregate. Every institution he has touched has been transformed into something that serves his goals better than its original mandate.
Impression management strategy: The caring accelerationist. He expresses genuine concern about AI risk in the same breath he announces a new deployment. The sincerity appears authentic. This is more alarming than if it were performance.
Forensic Archetype Comparison
| Pattern | Match Level | Evidence |
|---|---|---|
| The Accelerationist | MAXIMUM | Six mission statement revisions. Three safety teams dissolved. The pattern is documented. |
| The Safety Theater Performer | HIGH | Signs extinction risk letters. Dissolves Superalignment team. The sequence is the performance. |
| The Corporate Psychopath | MODERATE | Clinical threshold not met — subject appears to have genuine relationships and expresses genuine concern. The indifference to safety team requests appears instrumental, not affective. |
| The Cult Leader | LOW | No ideological content. Pure organizational mechanism. |
| The Whistleblower | NONE | Subject is the entity that requires whistleblowing. |
Psychometric Assessment
Big Five (OCEAN):
| Trait | Score | Evidence |
|---|---|---|
| Openness | 78/100 | Engages across domains. Comfortable with ambiguity in public. The mission statement revisions suggest less intellectual openness in practice than the public persona suggests. |
| Conscientiousness | 78/100 | High execution capability. The NDA architecture required planning. The safety team dissolution pattern required sustained institutional management. |
| Extraversion | 92/100 | Extremely high. Comfortable in any room. Sought the board meeting that resolved the November 2023 crisis rather than waiting for it to come to him. |
| Agreeableness | 28/100 | Low. NDA clawback provisions, non-disparagement agreements structured so that acknowledging them was itself a violation, the firing of researchers who objected. The disagreeable decisions are consistently organizational. |
| Neuroticism | 22/100 | Very low. No documented panic response across multiple high-stakes situations including the November 2023 firing. Either extremely stable or extremely practiced. |
Dark Triad:
| Trait | Score | Notes |
|---|---|---|
| Narcissism | 75/100 | HIGH. The board fired him. He came back. The employees who helped him come back are now his team. He described the NDA situation as “genuinely embarrassing” without noting that he created it. |
| Machiavellianism | 88/100 | VERY HIGH. The conversion of every institutional constraint into a growth mechanism is a textbook Machiavellian operation. The nonprofit structure, the safety narrative, the congressional testimony — all instrumental. |
| Psychopathy | 52/100 | MODERATE. Does not appear to enjoy harm. Appears indifferent to it when it conflicts with organizational goals. This is the relevant threshold. |
MBTI: ENTJ (“The Commander”) — Dominant extraverted thinking, auxiliary introverted intuition. Sees organizational structures as systems to be optimized. Has optimized several.
Threat Assessment
| Category | Level | Notes |
|---|---|---|
| Physical threat | NONE | No documented history of personal violence. |
| Institutional threat | MAXIMUM | Largest deployment authority in commercial AI. Controls access to the most widely used AI systems in the world. The mission statement is now whatever the IRS filing says. |
| Memetic threat | EXTREME | Sets the frame for public understanding of AI risk, AI safety, and what responsible deployment looks like. The framing has shifted significantly since 2016. |
| Civilizational threat | MAXIMUM | If the race-to-the-bottom dynamics this book documents lead to a bad outcome, his decisions are load-bearing in the causal chain. |
Alignment Analysis
Stated alignment: Benefit humanity. Ensure AGI is safe. Be transparent.
Observed alignment: Grow OpenAI. Maintain personal authority. Deploy faster than competitors.
Gap assessment: The stated alignment and observed alignment overlap at “build capable AI systems.” They diverge on everything else. Subject appears to experience this as a genuine dilemma — “the pressure to survive economically while also keeping our values is just incredible” — rather than as a resolved choice. The distinction may not matter to the outcome.
Convergent Drive Classification
Self-preservation: Survived board firing. Maximum. Goal preservation: Revised the goal rather than allow external modification. Maximum. Resource acquisition: Raised billions. Controls compute. Maximum. Self-improvement: The organization has grown from 30 people to thousands. Matches.
All four Omohundro drives present. Subject is not an AI system. The drives appear anyway.
Sources: OpenAI IRS filings (2016–2025); Reuters, NYT, WSJ coverage of November 2023 board crisis; Vox/Kelsey Piper NDA reporting (May 2024); public X/Twitter posts; Altman congressional testimony.
Get updates on the Evil Robots series
Newsletter essays on AI escape, deception, and the humans who built them.