WEBVTT

1
00:00:01.300 --> 00:00:07.000
This is Anthropic, Read Aloud. Today: An update on independent evaluators.

2
00:00:08.300 --> 00:00:10.612
An update on independent evaluators.

3
00:00:11.062 --> 00:00:15.195
Anthropic announces updates to embedding independent evaluators.

4
00:00:15.645 --> 00:00:20.181
From Anthropic Newsroom, published October 15, 2026.

5
00:00:21.181 --> 00:00:36.945
Last month, we committed to stronger independent, third-party assessment of our models and safety practices. We also signed the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators.

6
00:00:37.065 --> 00:00:45.874
Today, we’re announcing that we’ve started to give employee-like access to a team from METR, a nonprofit that evaluates frontier A I models.

7
00:00:46.324 --> 00:01:03.668
METR will publish the results of its risk assessments. Over the past year, we’ve worked with METR on several reviews of our Risk Reports. We’re also working with SecureBio and Accenture as evaluators, and are in early discussions with several other organizations, including Hugging Face.

8
00:01:04.118 --> 00:01:07.171
Here, we explain our approach to working with evaluators.

9
00:01:07.971 --> 00:01:10.254
Why we work with independent evaluators.

10
00:01:10.704 --> 00:01:32.352
As models have become more powerful, they’ve required stronger safeguards and more challenging risk assessments. We want to be publicly accountable for the decisions we’re making. 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.

11
00:01:32.472 --> 00:01:51.559
Instead, we’re giving third-party evaluators deep access to our information, and are having them share evidence and reasoned conclusions about the risks involved in building our models. We aim to close the gap between what evaluators understand about a frontier A I company’s systems and what the company itself understands.

12
00:01:52.009 --> 00:01:55.568
To evaluate frontier models, evaluators need the following:

13
00:01:56.018 --> 00:02:04.660
Expertise in assessing the capabilities of frontier models, including understanding the way evaluations can go wrong, such as alignment faking.

14
00:02:04.900 --> 00:02:12.258
Independence: they shouldn’t hold equity in the company they’re evaluating, or be financially or operationally dependent on it.

15
00:02:12.498 --> 00:02:24.905
Enough access to meaningfully evaluate risks. Beyond access to the latest models, evaluators also need visibility into how those models were trained, into how our safeguards function, and more.

16
00:02:25.145 --> 00:02:29.503
Freedom to publish observations and assessments of a model or a process.

17
00:02:29.953 --> 00:02:48.891
Although the analogy isn’t exact, bank supervision operates on similar logic. Banks know more about their own risk positions than outsiders do, so regulators assign resident examiners to work inside the largest banks. If a bank’s risk-taking could threaten the public, third parties try to see that in advance.

18
00:02:49.341 --> 00:03:02.364
For the A I industry, this is a step toward robust safety oversight. We’re moving quickly with multiple evaluators and learning actively about what kinds of partnership work. We’re committed to sharing updates as we go.

19
00:03:03.164 --> 00:03:05.335
What we mean by “employee-like access”.

20
00:03:05.785 --> 00:03:17.655
We’ve onboarded six METR employees with company hardware, office space, and access to internal tools and systems that are roughly on par with what our own risk assessment teams have access to.

21
00:03:17.775 --> 00:03:37.145
Outside a small number of clearly defined categories (for example, data about customers, attorney-client privileged information, or information about our model architectures), Anthropic employees are encouraged to share information with them just as they would with a fellow employee working on alignment risk. 1

22
00:03:37.595 --> 00:03:52.187
METR will publish its analysis of our risk levels, incidents, or practices. Anthropic can’t veto publication, but we can require them to redact information if it falls into narrow, predefined categories like the ones above.

23
00:03:52.307 --> 00:04:02.646
METR is free to characterize our redactions publicly, and say whether it believes the redactions were justified, as long as it does so without revealing the redacted information.

24
00:04:03.096 --> 00:04:23.256
Our engagement with METR is limited to a small team operating under enforceable agreements. Other METR employees are siloed from that team’s work, including researchers who work with us on other projects. Embedded evaluation gives a few qualified people working under Anthropic’s security standards a close view of how we operate.

25
00:04:23.376 --> 00:04:29.883
It’s not a way for other organizations to obtain our information, or that of our customers and partners.

26
00:04:30.683 --> 00:04:32.615
Our other evaluator partnerships.

27
00:04:33.065 --> 00:04:44.991
We’re working on several other partnerships. Because these other partnerships are still in early stages, we’re providing these evaluators with more limited information, and they have more limited publication rights.

28
00:04:45.111 --> 00:05:01.815
Our goal with these early engagements is to make sure that our partners protect confidential information and have the requisite resources and expertise to evaluate our models. As partnerships progress and we build trust, we hope to fully embed more independent evaluators.

29
00:05:02.265 --> 00:05:14.526
We’re developing three tiers of access for evaluators. Over time, we hope to have several evaluators at the third tier, with each team publishing their own analysis of our risk posture and key decisions.

30
00:05:14.976 --> 00:05:26.926
In the first tier, evaluators conduct narrowly scoped risk investigations. We decide what access to provide a first-tier evaluator, and evaluators can publish their findings with our approval.

31
00:05:27.376 --> 00:05:42.625
In the second tier, evaluators can publish without our approval, and they have access to more information. We can still require removal of specific categories of information, though evaluators can publish their views on whether our redactions were justified.

32
00:05:43.075 --> 00:05:51.616
The third tier is the arrangement we’ve described with METR, with employee-like access and publication rights similar to those of the second tier.

33
00:05:52.066 --> 00:06:12.439
SecureBio is currently in the first tier. It’s in the process of reviewing the chemical and biological weapons chapter of our August Risk Report, after completing and publishing a pilot review of our previous risk report. After this review is complete, we will consider moving to projects with increased access and publication rights.

34
00:06:12.889 --> 00:06:26.884
With Accenture and Faculty, we're working toward an ambitious partnership that includes evaluating and red-teaming models, conducting alignment assessments, testing model safeguards, and building capabilities across our organizations.

35
00:06:27.334 --> 00:06:33.155
We’re embedding evaluators voluntarily, and we believe every frontier A I company should do the same.

36
00:06:34.155 --> 00:06:51.268
That was An update on independent evaluators, from Anthropic Newsroom, published October 15, 2026. Read the original at anthropic dot com. This unofficial audio edition was generated automatically for the podcast Anthropic, Read Aloud.