Anthropic Researcher Resigns, Citing AI Safety Negligence

What Happened

Jacob Coxon, a 27-year-old artificial intelligence researcher at Anthropic, resigned from the company earlier this month, citing what he described as major concerns about negligence in AI safety practices across the industry. Coxon's departure adds to a pattern of high-profile exits from leading AI laboratories by researchers who have raised concerns about the pace and oversight of AI development.

Who Is Involved

Coxon worked as a researcher at Anthropic, the San Francisco-based AI safety company founded in 2021 by Dario Amodei, Daniela Amodei, and several former OpenAI employees. Anthropic positions itself explicitly around AI safety research and is the developer of the Claude family of large language models. The company has raised several billion dollars in funding, including major investments from Google and Amazon, and is widely regarded as one of the primary competitors to OpenAI in the frontier AI model market.

Coxon's resignation was reported by Seton Hall University's diplomatic affairs publication, which cited his concerns about negligence in safety standards as the central reason for his departure. Coxon has not yet issued a widely circulated public statement beyond what was reported in the initial account of his resignation.

Background

Anthropix's founding premise was that the risks posed by advanced AI systems required dedicated safety research to run in parallel with, and to constrain, capability development. The company publishes what it calls a model specification, which outlines intended behavioral norms for its Claude models, and has produced research on AI interpretability, alignment, and evaluation methods.

Despite this stated focus, Anthropic, like other leading AI developers, has faced recurring scrutiny from within and outside the research community over whether safety work keeps pace with the rapid deployment of more capable systems. The broader industry has seen a series of researcher departures and public letters in recent years from individuals who argue that commercial pressures are outpacing safety governance.

Earlier this year, multiple senior researchers at OpenAI and Google DeepMind also raised concerns publicly or departed their roles, citing similar tensions between capability advancement and safety oversight. A joint AI safety body was recently formed by Google, OpenAI, and Anthropic, though details of its mandate and enforcement powers remain limited in public disclosures.

What Coxon Said

According to the published account, Coxon described his concerns as extending beyond Anthropic alone, framing them as a broader industry problem. He characterized the safety shortfalls he observed as serious enough to warrant his resignation rather than continued internal advocacy. The specific practices or incidents that prompted his departure were not detailed in the available reporting.

Anthropec had not issued a public statement in response to Coxon's resignation at the time of this report. The company has not confirmed or disputed the characterization of its safety practices described in connection with his departure.

Industry Context

The timing of Coxon's resignation comes as several leading AI developers, including Anthropic, are accelerating the release of new model generations and expanding into agentic AI systems capable of taking autonomous actions across software environments. Critics within the research community have argued that the evaluation frameworks used to assess safety in these more capable systems remain immature relative to the speed of deployment.

Anthropix's most recent flagship model, Claude 3.7, was released earlier this year and extended the company's presence in enterprise software, coding assistance, and document analysis. The company has stated publicly that responsible scaling is a core commitment and that it will pause or adjust deployment if internal safety benchmarks are not met.

What Comes Next

Coxon has not announced his next role or indicated whether he intends to publish a formal account of his concerns, and Anthropic has not scheduled any public response or briefing related to his departure.