← Accueil actu | Salle du marché

Anthropic pretraining researcher Jacob Coxon resigns, igniting broad AI safety debate

2026-09-10 · Source:SiliconANGLE

Anthropic pretraining researcher Jacob Coxon resigns, igniting broad AI safety debate

Jacob Coxon left Anthropic's AI pretraining team, warning in X posts viewed millions of times that AI labs are 'gambling with our lives.'

Jacob Coxon, a researcher on Anthropic PBC's AI pretraining team, resigned from the company on September 9, 2026, announcing the decision in a series of posts on X that have been viewed millions of times. Coxon remained part of the pretraining group until the day of his announcement, and the exit quickly drew widespread attention.

In the posts, Coxon wrote that his decision was motivated by concerns about the direction of artificial intelligence development, warning that AI laboratories are 'gambling with our lives.' The blunt framing, delivered from inside the pretraining effort at Anthropic itself, helped the announcement spread rapidly, with the posts accumulating views in the millions as they were shared and debated across the platform.

The resignation has since sparked a broad discussion about AI safety, with the conversation extending well beyond Coxon's own account as observers weighed what the departure of a pretraining researcher might signal about sentiment inside AI companies. Pretraining — the stage in which large models are trained on massive datasets — sits at the core of how companies like Anthropic build their systems, giving the departure added weight in an industry where talent movements are closely watched.

Anthropic is organized as a public benefit corporation, a structure that formally ties the company's mission to public benefit alongside commercial aims, and the episode lands amid ongoing debate over how AI laboratories balance rapid development against safety commitments. Coxon's posts, amplified by millions of views on X, turned a single resignation into a talking point on whether the industry is proceeding responsibly — a question he framed in stark terms: 'gambling with our lives.'

Mots-clés: Anthropic · Jacob Coxon · AI Safety · Pretraining


À lire aussi


萬安算交所 AIXX · AIXX · Bourse compute vérifiée · Accéder à la salle