---
title: Anthropic Partners with METR and SecureBio to Evaluate AI Risks
url: https://www.dataloco.com/en/anthropic-partners-with-metr-and-securebio-to-evaluate-ai-risks
published: 2026-10-22T20:21:20+00:00
language: en
section: AI
source: https://itforum.com.br/noticias/anthropic-firma-parceria-com-metr-e-securebio/
organizations: Anthropic, METR, SecureBio, Hugging Face, Accenture, Nvidia
publisher: Dataloco
---

# Anthropic Partners with METR and SecureBio to Evaluate AI Risks

Anthropic has established a partnership with the nonprofit organizations METR and SecureBio to assess the company's security practices and risk management. This move integrates independent evaluators into the company's operations to analyze the development and safety of its artificial intelligence models. The announcement was made on Thursday, October 15, 2026, marking a significant step in the company's approach to external oversight.

The initiative follows a proposal published by Anthropic's chief executive officer, Dario Amodei, in the previous month. In that proposal, Amodei suggested slowing the pace of AI model development. Following this proposal, Anthropic selected Accenture as its first external partner. The company is also in negotiations with other organizations, including Hugging Face, to expand its program of embedded independent evaluators. The partnership with METR and SecureBio is part of a broader commitment to incorporate external scrutiny into the company's development process.

Anthropic has created a three-tier system for these evaluators, each with a distinct level of access to the company. First-level evaluators conduct risk investigations with a restricted scope. Second-level evaluators are permitted to publish reports without prior approval from Anthropic. Third-level evaluators have access equivalent to that of internal employees. METR falls into the third category. Six professionals from METR have been integrated into Anthropic and have access to the company's equipment, offices, and tools at a level comparable to internal risk evaluation teams. These professionals will analyze risk levels, incidents, and company practices, and they will publish their conclusions. Anthropic does not fund the work of METR.

SecureBio is currently reviewing the section of Anthropic's risk report concerning chemical and biological weapons, which was published in August 2026. The organization may expand its participation in projects with greater access and publication rights in the future. The level of access that may be granted to Hugging Face has not yet been defined. Hugging Face, known for operating an open-source code development platform, was acquired by Nvidia for approximately 13 billion US dollars in the month prior to the publication of this report.

The status of the partnership with Accenture remains open. Anthropic stated it is working toward an ambitious partnership with the consultancy but has not disclosed which level of the system it fits into. In the previous month, the company informed that it would directly fund Accenture's work, which generated criticism on social media regarding the consultancy's independence. Anthropic acknowledged the absence of consolidated standards for funding such models. The company stated that there are currently no parameters for what information embedded evaluators should access or how they should report their findings. There is also no established system for funding independent evaluation. Anthropic believes that in the long term, funding should come from collective or governmental sources, as advocated in its Advanced AI Framework in June. Since none of these sources exist today, the company plans to work with different evaluators under different funding arrangements.

Independent evaluators like METR occupied a discreet space in the AI industry before Amodei's proposal. With growing concerns about the ability of AI companies to protect their models, these organizations have gained relevance. The President of the United States, Donald Trump, encouraged companies in the sector to associate with an independent external auditor or evaluator in a voluntary agreement presented at the end of September 2026. Signatories of this agreement include executives from Google, Meta, OpenAI, SpaceX, and Nvidia. Anthropic stated that it is incorporating evaluators voluntarily and believes that all frontier AI companies should do the same.