Anthropic CEO Dario Amodei pushed back against criticism that his warnings about AI risks have fueled public backlash against the industry, responding to investor Gavin Baker, who argued on the All-In podcast and social media that Amodei’s cautionary messaging has contributed to growing opposition toward data centers and AI regulation. Amodei disputed the characterization that his messaging skews negative, saying his commentary has been balanced between risks and benefits, and noting he wrote his essay on AI’s potential specifically because he felt the industry wasn’t offering an inspiring enough vision. He argued that public distrust of AI stems from a broader, long-standing crisis of trust in corporations and government rather than from warnings issued by AI executives themselves, adding that the most legitimate criticism of companies like Anthropic is their failure to deliver tangible benefits so far. On regulation, Amodei rejected the notion that oversight necessarily concentrates power among large firms, saying Anthropic has deliberately crafted policy proposals designed to constrain frontier AI companies while supporting smaller competitors.
In other company news, Anthropic published a blog post addressing questions about its new text watermarking system for Claude, introduced to comply with the EU AI Act’s Transparency Code. The company said watermarking relies on subtle patterns in word choices that don’t affect output quality and are undetectable to readers without a decoding key, using the SynthID-Text approach developed by Google DeepMind. Anthropic said it plans to release a watermark detection API, and clarified that light editing likely won’t remove a watermark, while complete rewrites would. The company added that code will carry minimal watermarking due to its more constrained structure, and that other AI developers who signed the same EU code will implement comparable systems. The disclosures follow public debate and reports of some users canceling Claude subscriptions over the policy.
