Anthropic CEO Warns of AI Risks, $1.5B Copyright Settlement

At the Axios AI + DC Summit, Anthropic CEO Dario Amodei highlighted both the risks and benefits of AI, estimating a 25% chance of catastrophic outcomes and a 75% chance of highly positive results. Amodei clarified that “really, really badly” refers to scenarios capable of destabilising societal systems, posing existential threats, or severe AI misuse, not minor errors like autocorrect mistakes.

Anthropic AI Updates

He also warned that AI could replace up to half of all entry-level white-collar jobs and stressed concerns over the export of advanced chips to China, reflecting the economic and geopolitical stakes of AI development.

Anthropic settles major copyright lawsuitEarlier this month, Anthropic agreed to a $1.5 billion minimum payout plus interest to resolve a class action lawsuit by authors alleging the company illegally scraped millions of copyrighted books to train its AI models. Filed in San Francisco federal court, the settlement is one of the largest AI-related intellectual property agreements to date.

The deal covers claims from up to seven million authors, with initial payouts around $3,000 per book for 500,000 works. Anthropic has pledged to delete any disputed data. Legal experts described the settlement as unprecedented, sending a strong message against using copyrighted material from unlicensed sources.

Analysts noted that while the financial impact is significant, it represents less than 1% of Anthropic’s valuation, allowing the company to retain its models while complying with copyright laws.