Anthropic CEO admits mistakes, defends AI ‘race to the top’

Anthropic CEO Dario Amodei answered one of his company's more prominent safety critics on Tuesday night, October 20, in a reply on X of nearly 2,000 words. He conceded that Anthropic had mishandled some issues but argued that staying at the frontier of AI is the best way to make the technology safe, Business Insider reported. He was responding to Nathan Calvin, chief strategy officer of the advocacy nonprofit Encode AI. In a post of more than 1,500 words on Monday, Calvin listed nine actions he said showed Anthropic was racing toward superintelligence at the expense of its safety commitments. Those included compute deals worth more than $500 billion and a growing reliance on Claude to build its own models.businessinsider.com

A counter-list of costly choices

"Our strategy has always been what I've called the 'race to the top' — by showing how it's possible to build the technology responsibly, we can raise the bar for the entire industry," Amodei wrote. He also wrote that "any decision, on any topic, can always be portrayed as a nefarious scheme".indiatoday.in+1

He then set out his own nine decisions that he said were commercially costly but improved safety. They included supporting California's SB 53, keeping strict safeguards against biological misuse, and delaying the wider release of advanced models until stronger cybersecurity protections were ready. He also pointed to releasing the Mythos model only to a small group of cybersecurity partners and restricting sales to companies controlled by the Chinese state. He acknowledged that holding back frontier capabilities "had a broader cost," including slower progress in biology.businessinsider.com+1

The draft bill controversy

Amodei admitted that Anthropic could have handled some of Calvin's points better. One was a draft federal bill the company circulated in Congress this month, which critics said would make it harder for the government to block dangerous models. Under the draft, first reported by Politico, federal officials could stop a model only if it posed "a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people". An Anthropic spokesperson said the draft reflects the company's Advanced AI Framework.politico.com+3

Calvin also criticized Anthropic for changing its Responsible Scaling Policy. In February the company dropped a commitment to pause on its own if its safeguards fell behind its models' capabilities. "Why do the mistakes systematically tend to guard and favour Anthropic's interests?" Calvin asked.chaincatcher.com+1

Scrutiny ahead of an IPO

The exchange comes as Anthropic prepares for an initial public offering expected in November at a $2 trillion valuation. Last week the company reported third-quarter revenue of $16.8 billion.businessinsider.com

Amodei ended on a note of defiance. "We'll always try to do better but, again, I am extremely proud of our track record and stand by it," he wrote.businessinsider.com