Sam Altman says world should accept some bad things from AI in exchange for its benefits
In the debut episode of POLITICO's new daily newsletter and podcast Decoded, OpenAI CEO Sam Altman argued that the world should 'accept some bad things happening' in exchange for AI's massive benefits and personal autonomy, doubling down on his high-risk, high-reward stance and light-touch regulation preferences. He also explicitly rejected the idea of a single lab controlling powerful AI, publicly spotlighting his worldview split with Anthropic. Per Fortune, the remarks came after a contentious summer of OpenAI model safety controversies and industry voluntary safety commitments, drawing academic pushback over who decides acceptable risk, plus online mockery.
Confirmed
- Altman said AI may cause hacking, fraud and other harms, but society should tolerate them in exchange for broad benefits and personal freedom of use; open access and individual autonomy are worth some misuse risk because AI's good will far outweigh its bad. The full video was published on POLITICO's YouTube channel.
- Per @alexverem, Altman categorized fraud and data breaches as 'recoverable'; the same post noted an Italian bank lost €95 million to a voice-cloning scam that week as an ironic aside—not Altman's own claim.
- Asked about his worldview differences with Anthropic and Dario Amodei, Altman said he understands the position that AI is too powerful and dangerous to be controlled solely by one San Francisco lab deciding how benefits are distributed, but called it a 'completely unacceptable tradeoff' (per @markk, @ChrisGPT).
- Per @AuntieM123 citing The Guardian, this is Altman's latest statement on AI risk-benefit tradeoffs; @rohanpaulai, @luisdans, @cloneofsimo, @AaronBergman18 and @Afinetheorem relayed the same remarks.
- Per @fortune, the remarks came as OpenAI emerged from a contentious summer over model safety issues, with the industry signing voluntary safety commitments.
- Per @RobbWiller, Stanford scholar itaisher argued such risk tradeoffs should not be decided unilaterally by tech CEOs but through democratic processes.
- Per @PolarBearby's rebuttal, what should be sought is the overall interest of the planet 'including humans,' with AI not the sole measure of value.
- Per @CtrlAltDwayne, netizens mocked the remark with the Shrek line 'Some of you may die, but it's a sacrifice I am willing to make,' implying Altman was excusing AI's harms.
- After the remarks, Polymarket's 'US passes AI safety bill' market showed roughly 5% odds of passage by end of 2026 and 16% by mid-2027, with explicit rules on qualifying legislation.
Why it matters
- Altman's remarks put the 'open access vs. centralized control' divide on the table, naming the worldview split with Anthropic and likely intensifying debate over AI risk-benefit tradeoffs in industry and policy circles.
- Academics immediately questioned who gets to decide such tradeoffs, underscoring the governance dispute over acceptable risk.
- The timing is especially sensitive given OpenAI's unresolved model-safety controversies and the industry's freshly signed voluntary safety commitments.
- The 'recoverable losses' framing faced real-world pushback via the same-week €95M Italian bank voice-cloning fraud case, plus pop-culture memes.
- Prediction markets' low odds for AI safety legislation reflect pessimism about near-term US legislative action.
2026-10-05 ~ 2026-10-07 · 17 related posts
- Episode 1: Ex-OpenAI policy chief Brundage: OpenAI and Anthropic are essentially the same(2026-10-05, 7 posts)
- Episode 2: Sam Altman says world should accept some bad things from AI in exchange for its benefits(2026-10-05, 17 posts)
- Episode 3: Altman's 'Accept Some Bad' AI Remark Draws Pushback(2026-10-06, 2 posts)
- Episode 4: Guardian Slams Altman for Saying World Should Accept AI's Downsides(2026-10-06, 2 posts)
Primary sources
- Altman says world should accept some AI 'bad things' as industry signs Trump's voluntary safety pact — fortune ·
- Sam Altman: the world should accept some AI harms for the benefits and human agency — rohanpaul_ai ·
- Sam Altman rejects AI gatekeeping, distances himself from Anthropic on regulation — mark_k ·
- Sam Altman calls Anthropic's 'one lab should control AI' view 'a completely unacceptable tradeoff' — ChrisGPT · 2026-10-05
- Sam Altman says world must 'accept some bad things happening' for AI's benefits — Polymarket · 2026-10-05
- Polymarket prices US AI safety bill at just 5% by end of 2026 — Polymarket · 2026-10-05
- [source] Sam Altman rejects AI gatekeeping, distances himself from Anthropic on regulation — mark_k · 2026-10-05
- Sam Altman says world should accept 'bad things' from AI in exchange for its benefits — nordicinst · 2026-10-05
- Altman in POLITICO interview: "the world should accept some bad things" for AI's benefits — connoraxiotes · 2026-10-06
- Altman says world should 'accept some bad things' as AI scams cost a bank €95M in a week — alex_verem · 2026-10-06
- Altman says world should 'accept some bad things' for AI's benefits, mocked online — CtrlAltDwayne · 2026-10-06
- Sam Altman says world should 'accept some bad things' for AI's benefits, sparking pushback — PolarBearby · 2026-10-06
8 near-duplicate retellings: Auntie_M123 · luisdans · AaronBergman18 · cloneofsimo · Afinetheorem · RobbWiller · rohanpaul_ai · fortune