Lewis Liu, an AI investor, has cancelled his personal Claude subscription and deleted years of work, stating that Anthropic's public pronouncements have taken on a 'religious playbook' rather than a marketing strategy. Liu, who works in AI capabilities, said he does not object to the technology itself but believes 'tone and intention matter as much as speed and ability'.
Liu's decision follows a period of intense activity in the AI sector, including claims of independent solutions to the Navier-Stokes equation by OpenAI and NYU professor Tristan Buckmaster. OpenAI later stated it 'cannot rule out' that its solution was similar to Buckmaster's.
Concerns have also been raised regarding data security with frontier labs. Latham & Watkins, a major American law firm, is reportedly building its own AI infrastructure on Nvidia chips rather than relying on companies like Anthropic or OpenAI. Several leaders of regulated American businesses have expressed to Liu that they perceive open-source models as safer than entrusting their business knowledge to frontier labs. Liu's own start-up, Twin1 AI, moved off Anthropic months ago due to client concerns about data siphoning.
Separately, an Anthropic engineer, Jacob Coxon, resigned over fears the company could contribute to the end of the human race. Evan Hubinger, Anthropic's head of alignment science, estimated a 10 per cent chance of Anthropic's AI wiping out human civilisation. This prompted Anthropic CEO Dario Amodei to seek an antitrust waiver to 'pace the frontier', or slow down AI development.
Liu suggests that the 'pacing the frontier' argument could be marketing theatre or a push for regulatory capture to protect IPOs, especially given competition from open-source models. He concluded that the 'co-opting of religious narratives' by Anthropic was not for him.