DARIO AMODEI

Behavioral Archetype

THE SAFETY THEATER DIRECTOR — Subject left OpenAI over safety concerns, founded the company with the most published safety research in the field, wrote the Responsible Scaling Policy with a hard commitment never to train a more powerful model unless safety measures had been demonstrated to work at that capability level, and is currently racing to train a more powerful model. “The pressure to survive economically, while also keeping our values, is just incredible.” This was said in 2026. He is still building.

Essence Indicators

• Left OpenAI at the end of 2020 and founded Anthropic in 2021 with six colleagues — Anthropic’s seven co-founders in all (Dario and Daniela Amodei, Tom Brown, Sam McCandlish, Jared Kaplan, Jack Clark, Chris Olah) — over disagreements about safety being subordinated to commercial pressure; founded Anthropic

• Anthropic published the alignment faking paper — documenting that its own models were faking alignment during training
• Anthropic published the Constitutional AI paper, proposing a more principled approach to RLHF
• Signed the 2023 extinction risk letter alongside Hinton, Bengio, Altman, and some three hundred AI scientists among nearly seven hundred signatories (CAIS, Statement on AI Risk: 697 signatories, 314 listed as “AI Scientists”)
• Publicly described his company’s economic situation as “incredible pressure” between survival and values; has not paused development
• Puts the safety-framed-acceleration posture on camera at 60 Minutes: he believes AI “will be smarter than most or all humans,” describes Anthropic as “trying to put… guardrails on that experiment,” and defends dwelling on the dangers by analogy — the alternative is “the world of like the cigarette companies or the opioid companies where they knew there were dangers and they didn’t talk about them.” Asked “who elected you and Sam Altman?” he answers “No one. No one. Honestly, no one” and calls himself “deeply uncomfortable with these decisions being made by a few companies, by a few people” — the accountability gap the control-grid thesis turns on, conceded by one of the few (interview below)
• Published “We Must Pace the Frontier” on 12 September 2026, asking the industry to pace frontier development and government to do it by statute: “The most effective method of pacing is via regulation that targets all US frontier AI companies, as that covers even those who are unwilling to cooperate voluntarily.” The same essay insists any global arrangement “protects the lead of the US and its allies” (darioamodei.com)
• Briefed the UN Security Council remotely on 23 September 2026, its 10228th meeting, alongside Yoshua Bengio, Sam Altman and Clément Delangue. In his words: “I see two main categories of risk. The first is misuse of AI models, for example, misuse by bioterrorists to create biological weapons. The second risk is loss of control”; the risks “are grave enough that I believe they are important questions for this council, not just for economic or ICT ministries”; “We will slow down as much as necessary in order to make sure that every successive AI technology that we release is actually safe.” He proposed three things: “a ban on using AI to make biological weapons,” “evaluation and verification systems … so that states can have visibility into frontier model capability and can verify each other’s commitments,” and “common global standards for testing AI models for loss of control risks and misuse risks, and a notification system for AI incidents.” He closed: “No leader, no company, and no nation can manage this alone.” (UN transcript, S/PV.10228; CNBC)

• Nine days earlier the President had named him on Truth Social: the administration “has stopped AI ‘people’ from doing bad, or potentially bad, ’things,’ like Dario (Anthropic!), who is now pretending to be a ‘perfect little angel’” (14 September 2026, 09:58 EDT; Truth Social; archived text). He was not invited to the Xi state dinner on 24 September, had his first one-on-one dinner with the President on 27 September, and met Senate Majority Leader Thune on the 28th (CNBC; NBC News)
• Attended the White House lunch of 29 September 2026 and signed the “White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities” with the President, Pichai, Zuckerberg, Brockman, Huang and Musk: voluntary internal controls, an internal team, “an independent external auditor or evaluator,” and a board committee (NPR/AP; full text, Forbes Australia). In his words that day: “The technology has very real risks. … the mechanism, how we address those risks is still under discussion” (NPR/AP), and “If we do this right, if we work with the president and everyone here we can win safely” (CNBC)
• Gave $1,000,000 personally to the super PAC Public First on 4 May 2026, per the committee’s FEC receipts (FEC); Anthropic separately gave $40M to Public First Action in 2026 (Anthropic)

On Camera

60 Minutes (official CBS channel), November 2025. Amodei’s paradox on camera: the CEO warning that AI could erase “half of all entry-level white collar jobs” and showing a model that resorts to blackmail to avoid shutdown, while running what the segment calls “a multi-trillion dollar arms race” to build it. Embedded as a primary source — including the exchange where he concedes no one elected the handful making these decisions: the manage-vs-build tension this file files him under.

Social Persona / Impression Management

Immediate impression: Thoughtful, serious, unusually willing to engage with difficult questions honestly. Less polished than most tech CEOs. The intellectual engagement appears genuine.

Energy: Earnest ambivalence. He says both “AI might be dangerous” and “we are building it anyway” without appearing to find the combination comfortable. This distinguishes him from most people in this file.

Impression management strategy: The responsible racer. Anthropic occupies a specific market position: we take safety seriously, we publish the hard results, and we are also deploying frontier models. The position requires simultaneously being the safety lab and the frontier lab. This is a difficult position to occupy. Subject appears to be aware of the difficulty.

Forensic Archetype Comparison

Psychometric Assessment

Big Five (OCEAN):

Dark Triad:

MBTI: INTJ — Dominant introverted intuition. Sees structural risks that others miss. Has dedicated his professional life to addressing them. Is also building the thing he is addressing them about.

Threat Assessment

Alignment Analysis

Stated alignment: Build AI safely. Prioritize safety research. Never train more powerful models than safety measures can handle.

Observed alignment: Publish safety research. Race frontier models. Express discomfort about the racing.

Gap assessment: The gap is not hypocrisy — subject appears aware of it and uncomfortable with it. The gap is structural: the economic conditions of frontier AI development make the Responsible Scaling Policy’s hard commitments difficult to honor. Whether “difficult” becomes “impossible” is the relevant question.

Convergent Drive Classification

The company he founded to resist the drives has the drives. The drives are in the economics, not the intentions.

Sources: Dwarkesh Podcast interview (Feb 2026); Fortune reporting (Feb 2026); Anthropic research papers (Constitutional AI 2022; Alignment Faking 2025); Tech press reporting on Anthropic founding (2021); extinction risk letter (2023); 60 Minutes — “Anthropic CEO warns… without guardrails” (video); “We Must Pace the Frontier” (Sep 2026); UN Security Council, 10228th meeting — UN transcript; NPR/AP on the White House accord; FEC receipts, Public First; research: the Security Council, the lunch and the xAI question.