Anthropic is taking significant steps to address safety concerns as it prepares for its initial public offering, projected to exceed a remarkable $2 trillion valuation. In a recent investor presentation, the company highlighted its commitment to safety practices, aiming to boost investor confidence in its operations.
The move comes amidst growing scrutiny in the industry, especially following a security breach concerning models developed by OpenAI, which saw a hack of the popular platform Hugging Face. To bolster its safety measures, Anthropic has enlisted the help of the METR research group, known for its investigative role in the aforementioned incident. As part of this collaboration, Anthropic has integrated six METR team members into its operations, providing them with company hardware, access to internal tools, and office space.
This initiative aims to facilitate open communication between the METR evaluators and Anthropic employees, fostering a collaborative environment for risk assessment. Notably, Anthropic is not funding METR; the organization operates on donations from private individuals, philanthropic entities, and governmental bodies such as the AI Security Institute in the UK.
The evaluators will produce an analysis of their findings, which will be made public without Anthropic having the ability to block publication. However, certain sensitive information, such as client data and privileged legal communications, may be redacted at Anthropic’s request.
In addition to METR, Anthropic is working with Accenture, which was announced as a partner focused on safety testing. Both organizations have committed at least $1 billion to establish new safeguards in AI technology. Another organization, SecureBio, is also contributing by evaluating risks related to chemical and biological weapons within the company.
Anthropic expressed a belief that every leading AI firm should adopt a similar approach to embedding evaluators, emphasizing the industry-wide importance of transparency. METR has reiterated its readiness to collaborate with all AI developers to enhance public understanding of AI safety and risk.
© 2026 The Financial Times Ltd. All rights reserved. Not to be redistributed, copied, or modified in any way.
Editor’s Take
This initiative by Anthropic underscores the growing emphasis on accountability within the AI sector, especially as it gears up for public investment. The partnership with external watchdogs like METR could set a precedent for transparency, benefiting users and businesses by increasing trust. In a rapidly evolving landscape, such measures are essential for fostering a safe and responsible AI ecosystem.
Source: arstechnica.com
