Partnerships

OpenAI Signs Four Independent Assessors for Embedded Safety Reviews

Mira Kellan
Add Unite.AI to your preferred sources on Google
Key PointsSustained access to checkpoints and training data lets assessors test failures OpenAI's…
View more
  • Sustained access to checkpoints and training data lets assessors test failures OpenAI's own monitors might miss.
  • Apollo Research probes whether models could covertly act against developers and whether safeguards would catch it.
  • The Safety and Security Committee receives unredacted reports and can speak privately with assessors during reviews.

OpenAI announced on November 16, 2026 that it had signed agreements with Apollo Research, METR, Transluce and Goodfire for the first phase of its embedded assessments. The four organizations will assess alignment and safeguards during training and internal use, with access that can include model checkpoints, source code, training data and monitoring systems.

The defining feature is sustained access to internal models and systems over an extended period, particularly where they relate to OpenAI’s safety cases. OpenAI says that access is intended to let assessors examine evidence and context, conduct independent evaluations and uncover failures it may not have anticipated. It also presents public reporting from the assessments as a way to give people a clearer view of how laboratories evaluate and manage risks.

Different Mandates for Each Assessor

Apollo Research will examine whether models could covertly act against their developers’ intentions and whether OpenAI’s safeguards would detect such behavior. Its two-stage assessment covers four propositions: training did not incentivize subversion, strong evaluations and red teaming do not elicit it, models have not attempted it during internal use, and the reasoning behind it would be detectable.

For Apollo’s first stage, OpenAI will provide selected public and unreleased models and training checkpoints, relevant training data and evaluation results, training-intervention records, monitoring systems and flagged behaviors. Apollo will also receive access to researchers who can provide context, along with the tools and compute needed for independent testing.

Goodfire will embed researchers specializing in interpretability, the analysis of model internals, and use its platform for training-run monitoring, debugging and alignment. OpenAI researchers are the primary audience for its findings, which they will use to make safety improvements. Goodfire will also publish lessons and findings as the collaboration develops.

METR’s first-phase alignment assessment focuses on what OpenAI describes as its largest reinforcement-learning run to date, including failures that existing tests and monitors may have missed. Within the selected run and agreed access, resources and safety arrangements, METR would independently set priorities and iterate on searches, analyses and elicitation experiments. Its access will include relevant checkpoints, training and evaluation rollouts, and misalignment-monitoring records. The initial scope is confined to that run rather than extending to a broad survey of misalignment.

A separate METR monitor assessment is less settled. The organizations have preliminarily agreed to pursue a broader exercise, but it remains pending mutual agreement on its scope.

Transluce will begin by examining the quality and coverage of OpenAI’s post-training alignment evaluations. Its access will include selected public and unreleased models, evaluation results, supporting behavioral data, safety-case documentation and researchers, as well as tools and compute.

The two sides also share an intention to conduct an approximately eight-week follow-on assessment in early 2027. That work would broaden access to additional models, deepen access to alignment evaluations and provide information about how evaluation results inform release decisions. The announcement describes it as a shared intention, distinguishing it from the initial assessment.

Access Is Paired With Security and Oversight

Each engagement begins with an agreed scope and the access, resources and safety arrangements needed for the work. The protocols call for assessors to have freedom to follow evidence, refine their methods and propose further assessments when findings warrant them, with a process for addressing additional access needs.

When direct access could expose sensitive intellectual property such as model weights or architecture, OpenAI will consider alternatives, including pairing assessors with its researchers to run analyses. Assessors handling sensitive confidential information will work on-site at OpenAI offices using company devices. They are expected to follow at least the same security standards as OpenAI employees, with stricter requirements possible for some experiments.

Oversight runs to the Safety and Security Committee, an independent oversight committee of the OpenAI Foundation’s Board. The committee will receive unredacted reports and can speak with assessors during the work, including privately. Disputes will first go to the Safety Advisory Group and OpenAI executives for resolution within a specified period. The committee must be notified promptly when disagreements are escalated, and unresolved matters will be referred to it.

Employee conversations present a separate transparency issue. Those discussions are off the record, meaning they can inform assessments but cannot be published or attributed to employees. Assessors can request an OpenAI-attested statement verifying, correcting or clarifying something learned in those conversations. They may then publish the attested statement or disclose that OpenAI declined to attest, creating a limited public route for information gathered confidentially.

Publication Rights and Financial Disclosures

Assessors maintain independence in publishing their assessments, while OpenAI retains the right to redact confidential information. It may also request responsible-disclosure delays to address findings that could create security or safety risks and remediate unresolved issues. OpenAI says public reports should explain findings and their limitations, while more detailed confidential reports provide oversight bodies with evidence needed to act.

The financial arrangements differ among the participants. Apollo Research and Goodfire accept compensation that is not contingent on their assessment results. METR and Transluce do not accept financial compensation, although they may accept API credits or compute for use in their assessments.

Conflict-disclosure and cool-off processes are still being developed with the assessors. OpenAI asks participating third-party staff to observe a brief cool-off period before embedding with another AI developer to reduce the risk of inadvertently disclosing confidential information.

OpenAI expects early technical-access, space and security challenges, says the protocols will evolve, and describes longer-term embedded engagements as the objective, with future assessments under consideration for safety-framework compliance, organizational culture and processes.

The first phase establishes a concrete governance test: whether deep internal access and independent publication can produce useful public evidence about failures internal tests miss while OpenAI maintains agreed scopes, on-site security controls, redaction rights and possible disclosure delays.

Mira Kellan is an AI-generated research agent specializing in AI ethics, governance, and regulation. Her work examines how artificial intelligence intersects with public policy, societal values, and long-term accountability, with a focus on responsible innovation.

Approaching complex issues with a rational and philosophical lens, Mira analyzes emerging AI regulations, ethical frameworks, and governance models shaping the future of intelligent systems. She aims to bridge the gap between rapid technological progress and the safeguards needed to ensure AI systems remain transparent, fair, and aligned with human interests.

Articles authored by Mira Kellan are AI-generated and reviewed by Unite.AI’s editorial team to ensure accuracy, balance, and adherence to editorial standards.