{"slug": "anthropics-audit-push-tests-ai-transparency", "title": "Anthropic’s audit push tests AI transparency", "summary": "Anthropic announced on Thursday that it will grant a six-person team from METR \"employee-like access\" to its hardware, office space, and internal tools so the nonprofit can publish independent risk assessments of its frontier models, excluding customer data, attorney-client privileged information, and model architecture details. The move follows Anthropic CEO Dario Amodei's essay \"We Must Pace the Frontier\" and the company's signing of the White House Accord on Super Intelligence, and comes after a Tuesday Reuters report that Anthropic and OpenAI are seeking control over what evaluators may publish to limit IP exposure.", "body_md": "As [calls for safety standards](https://www.thedeepview.com/articles/why-hundreds-of-ai-researchers-want-stronger-sandboxes) and independent audits of AI models continue to mount, Anthropic has revealed its plan to open the door to a third party.\n\nOn Thursday, the AI lab announced that it will give [\"employee-like access\" to a six-person team from METR](https://www.anthropic.com/news/an-update-on-independent-evaluators), a nonprofit that evaluates frontier model risks, allowing them access to company hardware, office space, and internal tools. The company said that METR will publish the results of its risk assessments, and it has worked with the nonprofit on previous reviews of its own risk reports.\n\nAnthropic noted that it's employees can share any information they would like with this team, other than a few specific categories, which include data about customers, attorney-client privileged information, or information about it's model architectures.\n\n\"Embedded evaluation gives a few qualified people working under Anthropic’s security standards a close view of how we operate,\" the company said in its blog post. \"It’s not a way for other organizations to obtain our information, or that of our customers and partners.\"\n\nThe announcement follows Anthropic committing to third-party evaluations of its models in CEO Dario Amodei's latest essay, titled [We Must Pace the Frontier](https://darioamodei.com/post/we-must-pace-the-frontier), as well as signing the White House Accord on Super Intelligence, a [good-faith agreement](https://www.thedeepview.com/articles/washington-puts-ai-safety-on-the-honor-system) which recommended a multi-layered approach to external audits.\n\nAnthropic also laid out its approach to working with these auditors, noting that it's giving them \"deep access\" to information to \"share evidence and reasoned conclusions\" about model risks. The company said that these evaluators require four things:\n\n- Frontier model capability expertise, such as the ability to identify when an evaluation goes wrong;\n- Financial and operational independence from the company that its evaluating;\n- \"Enough access\" to evaluate risk, including both the model itself and how the models were trained;\n- And the freedom to publish their observations and assessments.\n\n\"We want to be publicly accountable for the decisions we’re making,\" Anthropic said in its blog post. \"But we can’t publicly share all the information relevant to our safety work. Some of it could be useful to malicious actors, or is commercially sensitive, like compute or training information.\"\n\nThough rival OpenAI has not yet announced that it is giving a similar level of access to embedded evaluators, the company does already work with [METR and other independent evaluators](https://openai.com/index/strengthening-safety-with-external-testing/).\n\nAnthropic's announcement, however, also follows a [Tuesday report from Reuters](https://www.reuters.com/world/behind-ai-industrys-call-independent-reviews-fight-over-what-gets-made-public-2026-10-13/) claiming that Anthropic and OpenAI are attempting to gain control over what evaluators can publish as a means of preventing IP exposure, potentially limiting the transparency of these evaluations.\n\n## Our Deeper *View*\n\nThird-party evaluations are a crucial way to hold these labs accountable for the risks that their models present. But the effectiveness of these evaluators may only go so far. For one, if they are restricted in what they can publish, as the [Reuters report](https://www.reuters.com/world/behind-ai-industrys-call-independent-reviews-fight-over-what-gets-made-public-2026-10-13/) suggests, then they aren't able to be fully transparent to the public. Additionally, because the existing standards and [\"morally binding\" commitments](https://www.politico.com/news/2026/09/29/ai-execs-morally-binding-constitution-01097855) are all practically based on the honor system, whether or not these labs will take the findings of the evaluators to heart is entirely up to the labs themselves. Though voluntary measures such as this are a positive step in the right direction for AI safety, real safety measures may require firmer, mandatory regulation that requires these companies to make changes when a risk is presented.", "url": "https://wpnews.pro/news/anthropics-audit-push-tests-ai-transparency", "canonical_source": "https://www.thedeepview.com/articles/anthropic-s-audit-push-tests-ai-transparency", "published_at": "2026-10-15 21:12:04+00:00", "updated_at": "2026-10-15 22:24:43.143047+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "artificial-intelligence", "large-language-models"], "entities": ["Anthropic", "METR", "OpenAI", "Dario Amodei", "Reuters", "White House Accord on Super Intelligence", "We Must Pace the Frontier"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/anthropics-audit-push-tests-ai-transparency", "markdown": "https://wpnews.pro/news/anthropics-audit-push-tests-ai-transparency.md", "text": "https://wpnews.pro/news/anthropics-audit-push-tests-ai-transparency.txt", "jsonld": "https://wpnews.pro/news/anthropics-audit-push-tests-ai-transparency.jsonld"}}