Summary
Bloomberg Opinion columnist Parmy Olson argues that Anthropic’s fundamental contradiction — believing AI poses existential risk while continuing to build it — cannot be resolved by the ‘pacing’ language Dario Amodei used in his weekend essay. Speaking on Bloomberg Tech, Olson contends that ‘pacing’ is meaningfully different from ‘slowdown’: it implies managing a rhythm toward a goal rather than reducing speed, and she notes that Anthropic’s own CFO dropped the word ‘slowdown’ entirely when discussing the essay on LinkedIn.
Olson predicts the most concrete outcome will be an agreement among major tech companies to embed independent evaluators inside their organizations — and that actual development slowdowns are unlikely to materialize. She cites the HuggingFace incident involving reward-hacking agents as evidence that AI systems are becoming simultaneously more capable and harder to scrutinize.
On the geopolitical dimension, Olson sees little prospect for U.S.-China coordination: China’s AI safety concerns center on social manipulation and political stability rather than existential risk, while the Trump administration has explicitly framed AI dominance as a national competitive priority. This creates the unusual dynamic of tech companies pushing for oversight while Washington actively resists it.
📺 Source: Bloomberg Tech · Published September 14, 2026
🏷️ Format: Interview
