Image credit: Shutterstock.
Key Takeaways
- Amodei defended his AI risk messaging, saying his essays split evenly between risks and benefits despite a perception of doom-heavy framing.
- Musk called the exchange “interesting” and said he hopes AI “is nice to us,” without addressing specific arguments.
- Amodei argued for mandatory frontier-model testing and pointed to California’s SB 53 to counter claims that regulation favors big AI firms.
Elon Musk said early Sunday that he hopes artificial intelligence treats humanity well, commenting on a thread in which Anthropic CEO Dario Amodei defended his public warnings about AI risk. Musk’s post came a minute after he called the underlying exchange “interesting” in a reply to Amodei’s thread.
What Started the Exchange
Investor Gavin Baker opened the discussion, arguing that rumors about Amodei circulate easily because they match the Anthropic CEO’s public tone. Amodei responded directly in a lengthy thread on X, rejecting the idea that his messaging skews mostly toward danger.
Amodei said his two major essays on AI split evenly between risks and benefits. He argued that short interview clips create a distorted impression because negative framing draws more attention than balanced discussion.
Amodei said he wrote “Machines of Loving Grace,” his essay outlining AI’s potential benefits, partly because he believed the industry had painted too bleak a picture of the technology’s trajectory.
Amodei also pointed to a broader cause behind the skepticism his statements draw. He said ordinary people already distrust companies, governments, and the technology industry generally, and that AI inherits that suspicion rather than creating it on its own.
He framed his own criticism as a symptom of that wider trend rather than a unique reaction to Anthropic specifically.
Musk’s Response and the Broader Debate
Musk flagged the thread as “interesting” before posting his own reaction roughly a minute later. His follow-up post read:
“I hope AI is nice to us.”
The comment echoed a separate sentiment from investor Naval Ravikant, who has argued that once a system reaches superintelligence, humans lose the ability to guarantee control over it, leaving benevolence rather than oversight as the remaining safeguard.
Musk has spoken favorably about Anthropic in the past. In July, he criticized OpenAI for moving away from open-source development while pointing to Anthropic’s approach as a contrast. The exchange did not include any indication that Musk and Amodei disagreed on substance. Musk’s post did not address specific arguments raised by either side in the thread.
Amodei’s Case for Regulation Over Voluntary Pledges
Amodei used part of the thread to argue against the idea that regulation automatically concentrates power among the largest AI companies, a criticism frequently raised by smaller AI firms and open-source advocates.
He pointed to California’s SB 53, an AI safety law Anthropic supported, noting that it exempts companies with less than $500 million in annual revenue, which he said undercuts the claim that safety rules primarily protect incumbents.
The comments extend a policy position Amodei has pressed since June, when he called for mandatory testing of frontier AI models rather than reliance on voluntary industry commitments.
He later used an appeal for G7 coordination to warn against a fragmented global regulatory approach to AI, arguing that inconsistent national rules could push development toward jurisdictions with the weakest oversight.
In the thread, Amodei named two testing frameworks he supports: a reported Trump administration plan to test frontier models before public release, and a FINRA-style self-regulatory watchdog proposed by Google DeepMind CEO Demis Hassabis.
Both models would introduce a checkpoint before a model reaches the public, though neither has been formally adopted as U.S. policy.
Amodei’s Disease-Cure Timeline and the Criticism He Accepts
Amodei also addressed criticism of Anthropic’s public promises directly.
“I think by far the most accurate criticism of AI companies including Anthropic is that we haven’t yet delivered on our big promises to benefit the world.”
He returned to what he has described as his boldest prediction: that AI applied to biology could cure most human diseases within five to 10 years. Amodei has tied that timeline to his father’s death from Hepatitis C, which came shortly before curative antiviral treatments became available to patients.
Anthropic has positioned AI as core infrastructure in healthcare work with Pfizer, part of a broader push by the company to apply its models to drug discovery and biological research.
Amodei said he now views generic claims that AI will cure cancer as a cliché rather than a meaningful promise, arguing that only the outcome itself, not the marketing around it, will settle the question.
