ai-safety regulations ai impact-on-labor ai-governance policy ai-regulation ethics

Share this with friends, journalists and politicians

AI labs should give independent evaluators employee-like access

Key insights

Across researchers, executives, lawmakers, and government bodies, a common distinction is that evaluating public-facing models is not the same as evaluating how companies build and use AI internally.
Many supporters tie employee-like access to publication rights and conflict-of-interest safeguards, suggesting that access alone is not enough to establish credibility.
Critics and cautious supporters focus on many of the same practical questions—who chooses, pays, and supervises evaluators—even though they draw different conclusions from them.
Several quotes treat current company commitments as pilots or first steps and argue that standards or law would be needed to make the practice durable and universal.
A recurring tension is between giving evaluators very broad technical access and preserving protections for customer, third-party, or legally sensitive information.

Arguments for

Internal risk visibility

Proponents say many important risks arise during training, internal deployment, and other non-public internal uses that ordinary API-based or one-off external testing cannot see. Embedding evaluators with ongoing employee-like access would let them inspect internal systems, staff, documentation, and incidents as they happen.

Require independent embedded evaluations to close the internal deployment gap. Qualified independent third parties should test models throughout development and internal use, with employee-equivalent access.

Marius Hobbhahn · Sep 30, 2026 · full quote
New work from our team makes the case for embedded assessments — where independent evaluators get employee-like access to an AI company’s internal systems to help scrutinize and reduce the risk from internal AI use.

Centre for the Governance of AI (GovAI) · Sep 25, 2026 · full quote
Third-party evaluators mostly test frontier AI models through an API. That misses the risks that come from how companies build and use AI internally. [...] Independent evaluators get employee-like access to a company’s internal systems, staff, and documentation. [...] What do we …

Jonas Freund · Sep 2026 · full quote
Each frontier AI company commits to giving ongoing, employee-like access to a team of embedded third-party evaluators (such as METR), whose role is to verify adherence to safety practices and commitments, report incidents, and help assess the alignment of not just completed AI mo…

Dario Amodei · Sep 12, 2026 · full quote
Frontier AI labs need embedded third-party evaluators with ongoing, employee-like access. Not periodic audits. Not company-written safety reports. People inside the building who can independently verify what’s actually happening in training and deployment.

Joel Ramkhelawan · Sep 2026 · full quote
AI companies should be able to answer some very basic questions about their training process, such as: Did the AI ever actively try to undermine its own alignment training while it was going through the training? The answer to this should be an unequivocal no, and right now we ar…

Alexander Meinke · Sep 16, 2026 · full quote

External accountability

Advocates argue that the public should not have to rely on labs to mark their own work, especially when the technology is powerful, fast-moving, and not fully understood even by its creators. Independent embedded evaluators are presented as a way to verify safety claims, add more eyes to the problem, and provide concrete oversight and accountability.

I think Australians and anyone in the world is entitled to demand an independent oversight of these companies because they are powerful, because their technology is evolving very quickly and because they themselves say they don't have a complete handle on what these models are do…

James Paterson · Sep 28, 2026 · full quote
Dario Amodei proposed that frontier AI labs give independent evaluators ongoing, employee-like access to examine safety practices, report incidents and publish their own findings. [...] We think they are right. In a high-stakes field, trust should not rest on a company marking it…

Lawmence Wong · Sep 13, 2026 · full quote
Third-party experts — granted real access to the most capable frontier models — independently evaluate the risks, review the companies' own claims, and publish whether they agree. It's the difference between taking a company's word for it and having independent experts actually c…

Fathom · Aug 2026 · full quote
Evaluation by independent third parties will allow frontier AI organisations to draw on external expertise, have more ‘eyes on the problem’, and provide for more accountability.

Government of the United Kingdom · Oct 27, 2023 · full quote
Frontier AI development should move at the pace of our ability to keep it safe, not just what’s technically possible. The OpenAI–Hugging Face incident earlier this year proved that to keep AI safe, we need visibility into what’s happening within leading AI companies. As developer…

Brad Carson · Sep 12, 2026 · full quote
The public should demand transparency about what is happening inside AI labs: companies should disclose more information about the pace of capability gains and their progress towards recursive self-improvement; there should be stricter reporting requirements for safety incidents …

Joe Benton · Sep 11, 2026 · full quote

Editorial independence

Supporters stress that access only helps if evaluators remain meaningfully independent, control their own conclusions, and can publish findings without company editorial control, subject only to narrow redactions for sensitive or privileged material. They present these protections as necessary so embedded review answers beyond the company rather than becoming a managed narrative.

Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators access equivalent to that of their…

Percy Liang · Sep 18, 2026 · full quote
Two leading AI companies have just committed to giving outside evaluators a seat inside the building. On September 12, Anthropic’s Dario Amodei committed to giving independent evaluators “ongoing, employee-like access” to verify the company’s safety practices, with the right to p…

Elham Tabassi · Sep 24, 2026 · full quote
The one idea worth holding the industry to is embedded evaluators with independent publication rights. Give third-party safety researchers permanent access inside the labs, comparable to what bank examiners have inside banks, and let them publish what they find without the compan…

Denis Calderone · Sep 14, 2026 · full quote
Anthropic and OpenAI committed to providing independent third-party evaluators with employee-level access to assess the security and safety risks associated with model development and behavior. They have also said they will allow reviewers to publish findings without the companie…

Vinh X. Nguyen · Sep 14, 2026 · full quote
Amodei’s best idea is to give outside evaluators employee-level access to the labs. That only works if the evaluators answer to the public.

Imran Khalid · Sep 20, 2026 · full quote
Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators access equivalent to that of their…

Marietje Schaake · Sep 18, 2026 · full quote

Access parity

Many supporters argue that trustworthy evaluation requires access comparable to highly privileged internal employees, including relevant systems, data, tools, personnel, physical spaces, model versions, fine-tuning interfaces, and reduced-safeguard configurations when needed. They pair this with explicit exceptions to protect customer and other third-party sensitive data.

Frontier AI companies should grant embedded evaluators access equivalent to that of their own highly privileged employees for the purposes of their evaluations, and with exceptions to protect sensitive data belonging to the company’s customers and other third parties. This includ…

Stuart J. Russell · Sep 18, 2026 · full quote
A large frontier developer shall provide the independent safety evaluator access reasonably necessary to conduct a meaningful technical evaluation, including, where reasonably necessary, access to relevant model checkpoints, interfaces, documentation, testing environments, fine-t…

Patricia Fahy · Sep 18, 2026 · full quote
Apply access parity between internal and external evaluators. Where a frontier AI company’s internal evaluations relied on particular access modalities, then their external evaluators should receive functionally equivalent access unless the AI company can demonstrate this is unne…

Apollo Research · May 20, 2026 · full quote
It is clear from our experience that to run high quality evaluations that elicit high fidelity information to the potential risks posed by frontier systems, it is important for independent evaluators to have: Access to a Helpful Only (HO) version of the model, alongside the Helpf…

AI Security Institute (AISI) · Oct 2024 · full quote
Evaluators cannot conduct a trustworthy evaluation without sufficient access to the systems and information in question, for a reasonable amount of time. While there is a legitimate need for AI providers and evaluators to safeguard intellectual property and user privacy, some sys…

AI Evaluator Forum · 2025 · full quote
Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators access equivalent to that of their…

Daniel Kokotajlo · Sep 18, 2026 · full quote

Durable requirements

Several backers say voluntary commitments and pilot programs are promising but not durable enough on their own. They call for legal requirements, standards, and policy structures so embedded evaluation becomes universal and enforceable rather than something a company can abandon when inconvenient.

Bringing in independent evaluators is a real step. But a voluntary commitment can be dropped the second it gets inconvenient. The FRONTIER Act makes safety and transparency the law, and puts independent auditors in place to hold frontier labs to it. A promise you can break isn't …

Lori Trahan · Sep 13, 2026 · full quote
Permanent, employee-level access for independent evaluators should be required by law. We will join Anthropic in offering access to @METR before it’s required.

Eric Steinberger · Sep 12, 2026 · full quote
Require frontier AI companies to embed a designated independent verification organization onsite in their labs to conduct regular audits and evaluations.

Office of Governor Gavin Newsom · Sep 18, 2026 · full quote
Moving embedded auditing from a voluntary practice to an effective and universal industry requirement requires not only pilot projects, but also standards, tooling, and thoughtful policy design.

AI Verification and Evaluation Research Institute · Sep 13, 2026 · full quote
Article 92(1) of Regulation (EU) 2024/1689 empowers the Commission to conduct evaluations of general-purpose artificial intelligence (‘AI’) models. Article 92(2) of Regulation (EU) 2024/1689 empowers the Commission to appoint independent experts to carry out evaluations on its be…

European Commission · Jul 20, 2026 · full quote
I think Australians and anyone in the world is entitled to demand an independent oversight of these companies because they are powerful, because their technology is evolving very quickly and because they themselves say they don't have a complete handle on what these models are do…

James Paterson · Sep 28, 2026 · full quote

Arguments against

Capture concerns

Opponents argue that evaluators who work inside a lab, use its systems, or are chosen and paid by it may become too close to the organization they are supposed to scrutinize. In this view, social familiarity and financial dependence can blunt criticism and undermine any claim of true independence.

They really become part of the company, which makes it difficult to criticize it because the staffers become almost your friends.

Lilian Edwards · Sep 17, 2026 · full quote
'independent' evaluators paid for by the industry they evaluate seems like a recipe for tyranny

Vega (@vega_holdings) · Sep 12, 2026 · full quote
Third party evaluators that they choose is a cop-out. The evaluators need to be truly independent, with no ties to the labs at all. Preferably, people the labs hate and wish would leave.

Roon · Sep 12, 2026 · full quote

Strategic theater concerns

Some critics see the proposal less as robust safety oversight than as strategic damage control or moat-building by leading labs. Others question whether placing evaluators in offices is commensurate with the scale of the risks being invoked.

Yeah, nah. This is damage control. They're trying to get ahead of the recently riled up Joe Schmo who's worried AI might kill us now. They both have huge IPOs coming up and fears of an AI apocalypse could wreck their valuations. And they want to establish a procedure that would d…

Tumithak of the Corridors · Sep 12, 2026 · full quote
the bad news: Dario estimates that in 6 months AI will have the ability to take over the internet the worse news: Anthropic's current commitment to stopping this is... putting independent evaluators in desks in their office have a great day

tropoFarmer · Sep 12, 2026 · full quote

Middle ground

Implementation matters

A number of voices welcome embedded evaluators in principle but say their value turns on design choices such as who selects and funds them, how broad their access is, what they may publish, and whether serious findings can delay or block a release. On this view, employee-like access is necessary but not self-executing.

Its practical significance depends on the arrangements that follow: who chooses and funds evaluators, what they can inspect, what they can publish, and whether their findings can force a change when a lab disagrees.

Curtis Pyke · Sep 12, 2026 · full quote
Amodei proposed a generalized three-part plan of independent evaluators embedded inside frontier labs, coordination among companies, and global policy compatibility. [...] Implementing Amodei’s threshold idea for embedded evaluators, for instance, will require enforceable decisio…

Tom Wheeler · Sep 16, 2026 · full quote
We should take all the time we want to test things. In fact, I love this idea of having third-party testers. Oh, wow, I grew up in a company that's always done testing. So it's novel that we should say, wow, they're having embedded third-party testers. Right. Why not? It's a grea…

Satya Nadella · Sep 15, 2026 · full quote
Dario Amodei’s proposal to give independent evaluators employee-like access inside frontier AI laboratories addresses a real information problem. [...] Embedded evaluation needs a protected mandate, independent funding, broad technical and incident access, publication rights, and…

Zero Labs · Sep 12, 2026 · full quote

First step only

Some supporters describe employee-like access as a first step or one component of a broader safety architecture. They say it still needs to be paired with standards, tooling, public processes, coordination, or other measures before it counts as a full solution.

Excellent news! The details will matter, and this is a first step instead of a full solution, but a public commitment to employee-like access for independent evaluators from both frontier companies already puts us in a much better position than we were in yesterday.

Jeff Sebo · Sep 12, 2026 · full quote
Anthropic and OpenAI have now committed to embedded third-party evaluators with employee-like access to verify safety practices. We at SaferAI have been doing risk management work for years and welcome this development. [...] These are all pieces of one verification infrastructur…

SaferAI · Sep 17, 2026 · full quote
Frontier developers should engage at least one qualified independent evaluator to publish a review of their evaluations and risk reports. In parallel, governments and industry should develop the ecosystem of qualified independent evaluators by setting standards for them and ensur…

Anthropic · Jun 10, 2026 · full quote
I agree with Dario and am glad to see the support from leaders of the other frontier labs. The world needs a plan, developed in a global, public process. If we can slow down the growth of AI, we'll have more time to ensure the benefits outweigh the harm, and leave everyone better…

Bill Gates · Sep 13, 2026 · full quote
Dario's essay points towards the right path forward. The details need working through, but the direction is correct for meeting this critical moment. This is also why we recently put out our proposal for an industry-wide standards body for frontier AI.

Demis Hassabis · Sep 12, 2026 · full quote

The discussion is dominated by arguments for deeper third-party access inside frontier AI labs, especially to inspect internal development, deployment, and safety practices that outsiders cannot see through public interfaces alone. The main unresolved issues are how to keep embedded evaluators genuinely independent, what authority and publication rights they should have, and whether voluntary company commitments should be converted into binding standards or legal requirements.

Cast Your Vote:

Results (80 votes):

Total (80 votes)

For 74 (93%) Abstain 2 (3%) Against 4 (5%)

For (71)

Cloud Security Alliance AI Safety Initiative
Initiative of the Cloud Security Alliance focused on AI safety, security, and governance.

I am this person! Add a sourced quote Report comment

votes For and says:

Third, embedding addresses a real information problem. Frontier AI risk assessment depends on non-public facts such as training details, internal monitoring data, and incident records. An evaluator that can see only what the developer chooses to send...
more AI Verified source (Oct 17, 2026)

added 4d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Joint Select Committee on Artificial Intelligence
Bipartisan parliamentary committee of Australia appointed in 2026 to examine artificial intelligence opportunities, risks, governance, and national resilience.

I am this person! Add a sourced quote Report comment

votes For and says:

This would see independent, qualified and, over time, accredited individuals be embedded within the operations of the frontier labs, have employee-like access to their systems and activities to monitor and assist in managing risk.
AI Verified source (Oct 12, 2026)

added 3d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Marius Hobbhahn
CEO and co-founder of Apollo Research; AI safety researcher specializing in scheming and pre-deployment evaluations of frontier AI systems

I am this person! Add a sourced quote Report comment

votes For and says:

Require independent embedded evaluations to close the internal deployment gap. Qualified independent third parties should test models throughout development and internal use, with employee-equivalent access.
AI Verified source (Sep 30, 2026)

added 19d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
James Paterson
Australian Senator for Victoria and Shadow Minister for Defence.

I am this person! Add a sourced quote Report comment

votes For and says:

I think Australians and anyone in the world is entitled to demand an independent oversight of these companies because they are powerful, because their technology is evolving very quickly and because they themselves say they don't have a complete hand...
more AI Verified source (Sep 28, 2026)

added 24d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Centre for the Governance of AI (GovAI)
Nonprofit research organization helping institutions make better AI governance decisions as the world transitions to advanced AI.

I am this person! Add a sourced quote Report comment

votes For and says:

New work from our team makes the case for embedded assessments — where independent evaluators get employee-like access to an AI company’s internal systems to help scrutinize and reduce the risk from internal AI use.
AI Verified source (Sep 25, 2026)

added 23d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Elham Tabassi
Director of the AI and Emerging Technology Initiative at the Brookings Institution; former Chief AI Advisor at NIST and CTO of the U.S. AI Safety Institute

I am this person! Add a sourced quote Report comment

votes For and says:

Two leading AI companies have just committed to giving outside evaluators a seat inside the building. On September 12, Anthropic’s Dario Amodei committed to giving independent evaluators “ongoing, employee-like access” to verify the company’s safety ...
more AI Verified source (Sep 24, 2026)

added 25d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Gavin Newsom
Governor of California

I am this person! Add a sourced quote Report comment

votes For and says:

Proposals under consideration include requiring independent third parties to be embedded in frontier AI companies to verify safety frameworks and independently assess safety evaluations, as well as requiring companies to develop an emergency shutoff,...
more AI Verified source (Sep 23, 2026)

added 29d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Lama Ahmad
Technical program manager for Trustworthy AI at OpenAI, working on safe development and deployment of AI models.

I am this person! Add a sourced quote Report comment

votes For and says:

“OpenAI is committed to supporting independent assessments with deep levels of access across training, evaluation, and deployment. That access should enable assessors to challenge our assumptions, identify risks we may have missed, and reach their ow...
more AI Verified source (Sep 22, 2026)

added 25d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Mustafa Suleyman
Microsoft AI CEO; author

I am this person! Add a sourced quote Report comment

votes For and says:

The high level principle that has been agreed to universally across the industry is that we should invite independent oversight and independent audit. It's being called embedded evaluators. So these are people who have the technical expertise and the...
more AI Verified source (Sep 20, 2026)

added 10h ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Imran Khalid
Geostrategic analyst and senior fellow at Foreign Policy In Focus

I am this person! Add a sourced quote Report comment

votes For and says:

Amodei’s best idea is to give outside evaluators employee-level access to the labs. That only works if the evaluators answer to the public.
AI Verified source (Sep 20, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Patricia Fahy
New York State Senator for the 46th District and sponsor of the 2026 TERMINATOR Act on frontier-AI safety oversight.

I am this person! Add a sourced quote Report comment

votes For and says:

A large frontier developer shall provide the independent safety evaluator access reasonably necessary to conduct a meaningful technical evaluation, including, where reasonably necessary, access to relevant model checkpoints, interfaces, documentation...
more AI Verified source (Sep 18, 2026)

added 10d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Percy Liang
Stanford professor of computer science and director of the Center for Research on Foundation Models.

I am this person! Add a sourced quote Report comment

votes For and says:

Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators acce...
more AI Verified source (Sep 18, 2026)

added 15d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Daniel Kokotajlo
AI Futures Project founder

I am this person! Add a sourced quote Report comment

votes For and says:

Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators acce...
more AI Verified source (Sep 18, 2026)

added 16d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Accenture
Global professional-services company providing technology, consulting, and AI services.

I am this person! Add a sourced quote Report comment

votes For and says:

Accenture and Anthropic today announced that they are partnering to establish a team of embedded evaluators to work alongside Anthropic’s internal teams and safety partners to evaluate and red-team models, conduct alignment assessments, and test mode...
more AI Verified source (Sep 18, 2026)

added 19d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Daniel E. Ho
Stanford law and political science professor; senior fellow at Stanford HAI and director of the Regulation, Evaluation, and Governance Lab.

I am this person! Add a sourced quote Report comment

votes For and says:

We believe that all frontier AI companies should embed evaluators to independently assess AI risks, including evaluating the systems themselves and any significant incidents of real-world harm, as well as the companies’ training, deployment, oversigh...
more AI Verified source (Sep 18, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Marietje Schaake
Stanford Cyber Policy Center; former MEP; UN AI Advisory Body member; author of The Tech Coup

I am this person! Add a sourced quote Report comment

votes For and says:

Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators acce...
more AI Verified source (Sep 18, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Jacob Steinhardt
Associate Professor of Statistics and EECS at UC Berkeley; founder and CEO of Transluce, a nonprofit AI-oversight research lab.

I am this person! Add a sourced quote Report comment

votes For and says:

We believe that all frontier AI companies should embed evaluators to independently assess AI risks, including evaluating the systems themselves and any significant incidents of real-world harm, as well as the companies’ training, deployment, oversigh...
more AI Verified source (Sep 18, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Conrad Stosz
Head of Governance at Transluce; previously co-founded the U.S. Center for AI Standards and Innovation.

I am this person! Add a sourced quote Report comment

votes For and says:

We believe that all frontier AI companies should embed evaluators to independently assess AI risks, including evaluating the systems themselves and any significant incidents of real-world harm, as well as the companies’ training, deployment, oversigh...
more AI Verified source (Sep 18, 2026)

added 21d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Ben Buchanan
White House AI special advisor

I am this person! Add a sourced quote Report comment

votes For and says:

We believe that all frontier AI companies should embed evaluators to independently assess AI risks, including evaluating the systems themselves and any significant incidents of real-world harm, as well as the companies’ training, deployment, oversigh...
more AI Verified source (Sep 18, 2026)

added 24d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Joy Buolamwini
Algorithmic Justice League founder

I am this person! Add a sourced quote Report comment

votes For and says:

Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators acce...
more AI Verified source (Sep 18, 2026)

added 24d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Geoffrey Hinton
Godfather of Deep Learning

I am this person! Add a sourced quote Report comment

votes For and says:

Frontier AI companies should rely on evaluators that are meaningfully independent, that maintain full editorial control, and that disclose and mitigate potential conflicts of interest. [...] Frontier AI companies should grant embedded evaluators acce...
more AI Verified source (Sep 18, 2026)

added 24d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
David Duvenaud

I am this person! Add a sourced quote Report comment

votes For and says:

We believe that all frontier AI companies should embed evaluators to independently assess AI risks, including evaluating the systems themselves and any significant incidents of real-world harm, as well as the companies’ training, deployment, oversigh...
more AI Verified source (Sep 18, 2026)

added 24d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Arvind Narayanan
Princeton computer science professor

I am this person! Add a sourced quote Report comment

votes For and says:

We believe that all frontier AI companies should embed evaluators to independently assess AI risks, including evaluating the systems themselves and any significant incidents of real-world harm, as well as the companies’ training, deployment, oversigh...
more AI Verified source (Sep 18, 2026)

added 28d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Stuart J. Russell
AI Expert and Professor

I am this person! Add a sourced quote Report comment

votes For and says:

Frontier AI companies should grant embedded evaluators access equivalent to that of their own highly privileged employees for the purposes of their evaluations, and with exceptions to protect sensitive data belonging to the company’s customers and ot...
more AI Verified source (Sep 18, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Office of Governor Gavin Newsom
California governor's official press office

I am this person! Add a sourced quote Report comment

votes For and says:

Require frontier AI companies to embed a designated independent verification organization onsite in their labs to conduct regular audits and evaluations.
AI Verified source (Sep 18, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
SaferAI
Nonprofit organization researching and promoting responsible frontier AI risk management and safety governance.

I am this person! Add a sourced quote Report comment

votes For and says:

Anthropic and OpenAI have now committed to embedded third-party evaluators with employee-like access to verify safety practices. We at SaferAI have been doing risk management work for years and welcome this development. [...] These are all pieces of ...
more AI Verified source (Sep 17, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Alexander Meinke
Head of research at Apollo Research, an AI-safety organization focused on evaluating advanced AI systems.

I am this person! Add a sourced quote Report comment

votes For and says:

AI companies should be able to answer some very basic questions about their training process, such as: Did the AI ever actively try to undermine its own alignment training while it was going through the training? The answer to this should be an unequ...
more AI Verified source (Sep 16, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Tom Wheeler
Visiting Fellow at Brookings Institution; former Chairman of the Federal Communications Commission (FCC) 2013-2017; author and tech policy expert

I am this person! Add a sourced quote Report comment

votes For and says:

Amodei proposed a generalized three-part plan of independent evaluators embedded inside frontier labs, coordination among companies, and global policy compatibility. [...] Implementing Amodei’s threshold idea for embedded evaluators, for instance, wi...
more AI Verified source (Sep 16, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Satya Nadella
Microsoft CEO

I am this person! Add a sourced quote Report comment

votes For and says:

We should take all the time we want to test things. In fact, I love this idea of having third-party testers. Oh, wow, I grew up in a company that's always done testing. So it's novel that we should say, wow, they're having embedded third-party tester...
more AI Verified source (Sep 15, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate

Loading more opinions...

Abstain (2)

Loading more opinions...

Against (4)

Lilian Edwards
Internet law professor

I am this person! Add a sourced quote Report comment

votes Against and says:

They really become part of the company, which makes it difficult to criticize it because the staffers become almost your friends.
AI Verified source (Sep 17, 2026)

added 1mo ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate

Loading more opinions...

More h/ai-safety votes

AGI could make governments subordinate to the companies that develop it

Ban open source AI models capable of creating WMDs

AI labs should conduct risk assessments before high-stakes internal deployment of frontier models

[🇺🇸 Congress, Oct 2025] Requiring age verification and content safeguards for AI chatbots to protect minors

The U.S. and China should establish Mutually Assured Compute Destruction

Back