Leopold Aschenbrenner publishes 'Situational Awareness'
Aschenbrenner, dismissed from OpenAI's superalignment team months earlier for allegedly leaking information, argued the firing itself illustrated the security failures he described.
- Ideas & essays
- Major
Leopold Aschenbrenner, a former member of OpenAI’s superalignment team, published “Situational Awareness: The Decade Ahead,” a roughly 165-page essay arguing that artificial general intelligence was close, that its arrival would trigger a rapid jump to superintelligence, and that this trajectory made AI development a matter of national security comparable to the nuclear era. He projected AGI by 2027, extrapolating from the jump in capability between GPT-2 and GPT-4 and arguing publicly available information about compute and algorithmic trends was sufficient to see it coming — a claim of “situational awareness” he said only a few hundred people in AI labs and San Francisco genuinely possessed.
The essay argued that compute clusters would scale from roughly $10 billion to $100 billion and eventually to the trillion-dollar range, that this scale would require an entirely new industrial base including large increases in electricity generation, and that a race to superintelligence with China was likely to become the dominant frame for how governments treated frontier AI, potentially producing “national security forces not seen in half a century.”
Aschenbrenner wrote the essay months after OpenAI fired him, in April 2024, over what the company described as a leak; Aschenbrenner said he had shared a brainstorming document with external researchers for feedback and separately said his firing followed a memo he had sent OpenAI’s board warning that the company’s security was inadequate against espionage, particularly from China — a claim OpenAI disputed as the reason for his dismissal. That dispute gave the essay’s national-security argument a personal edge, and it circulated widely in Silicon Valley and policy circles; commentators divided over whether it was a serious extrapolation from public trends or an overconfident forecast dressed in the language of insider knowledge. Aschenbrenner went on to found an AI-focused investment fund built partly on the essay’s thesis.