The Diary Of A CEO
AI Whistleblower: We Are Being Gaslit By The AI Companies! They’re Hiding The Truth About AI
with Karen Hao
26 Mar 2026
28 min read
2h 2m
TL;DR
AI companies operate as "empires"—claiming intellectual property, exploiting labor, monopolizing research, and using existential risk narratives to justify anti-democratic development. Sam Altman manipulates narratives and people to advance OpenAI's agenda, while the industry gaslights the public by controlling which researchers get funded and censoring inconvenient findings.
Karen Hao is an investigative journalist who spent over 8 years covering the AI industry, including roles at MIT Technology Review. She interviewed over 250 people—90+ current and former OpenAI employees—to research her book "Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI." Her work exposes how AI companies operate like empires, claiming resources, exploiting labor, and monopolizing knowledge production while gaslighting the public about their true intentions.
Takeaways
1
**AI narrative capture enables undemocratic development** AI companies fund ~90% of AI researchers globally, setting agenda through money flows and censoring inconvenient findings (e.g., Google firing Timnit Gebru). The industry manufactures dual myths—utopian upside + existential downside—to justify why only they can safely develop AGI. This gaslighting strategy suppresses public participation in technology decisions that affect billions.
2
**Sam Altman's persuasion masks misalignment with allies** Altman convinced Elon Musk to co-found OpenAI by mirroring Musk's existential AI fears, then maneuvered to become CEO instead. Similarly, Dario Amodei (now CEO of competing Anthropic) felt manipulated into building systems contradicting his own values. Altman's success hinges on narrative elasticity—different definitions of AGI for Congress, consumers, and investors—not technical clarity.
3
**Why building human-replacement AGI is a choice, not destiny** The premise that AI = statistical models mimicking brains (Ilya Sutskever's hypothesis) drives decisions to scale models indefinitely. But this is unproven science, not consensus. Alternative approaches exist for focused AI systems (drug discovery, healthcare) without labor automation. Questioning "why AGI?" instead of accepting the industry's premise is essential before accepting the harmful side effects as inevitable.