PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Anthropic and OpenAI sought influence over what independent AI evaluators could publish, according to Reuters’ October 13, 2026 report. The account draws on unnamed people familiar with private negotiations and contracts Reuters says it reviewed; both companies offered qualifications that complicate the claim. The dispute is about where confidentiality ends and editorial independence begins: a lab may need to protect genuine trade secrets, but information it considers proprietary can also matter to judging a model’s risks.
What Reuters reported about the negotiations
Reuters reporter Aditya Soni wrote that two people familiar with private discussions said Anthropic and OpenAI had sought significant control over outside evaluators’ publications because the results might expose trade secrets. The sources were not named because the discussions were private. Reuters also said some existing contracts between labs and evaluators prohibit publishing confidential information without the company’s consent. These are claims attributed to sources and contract language Reuters said it reviewed, not a verified account of every lab’s terms.
The report described a short timeline: the sources said that as recently as October 5 the labs sought strict editorial control; by October 8, Anthropic’s talks with some evaluators appeared to have evolved. That account does not establish a final agreement or a settled policy across the industry.
How Anthropic and OpenAI described their positions
Anthropic
Anthropic said the report’s characterization did not accurately reflect all its discussions and work with independent evaluators. It said that in current conversations with evaluators it has worked with, reviewers control their methods, criteria, conclusions and what they publish. For new evaluator firms, Anthropic said it reserved a right to comment on and approve reports as the relationship is established.
#1 Best Overall
OpenAI
OpenAI said it was not seeking approval rights over full reports. It described redacting confidential information and said evaluators could disclose whether major redactions had been made. That distinction matters: a right to remove confidential material is not the same as control over a report’s findings, but readers need to know who decides what counts as confidential and how much has been removed.
Reuters reported that ideas under discussion included lab employees observing embedded evaluator teams and preventing evaluators working with one lab from simultaneously working with another. These were reported proposals, not established rules. Anthropic had named Faculty, Accenture’s specialist AI business, as its first embedded evaluator; Reuters said access and reporting details were still being worked out. Faculty CEO Marc Warner told Reuters: “Evaluators are only useful if people can trust what they report, so independence and transparency are critical.”
Why confidentiality can conflict with safety assessment
Some information may legitimately be sensitive: publishing it could expose proprietary techniques or other protected details. Yet the same information may help an evaluator determine what a system can do, how it was developed, or how a risk emerged. If a company can suppress or materially reshape findings by labeling them confidential, an evaluation can lose value even when the evaluator’s testing was rigorous.
Rumman Chowdhury, identified by Reuters as a former U.S. Science Envoy for AI and CEO of Humane Intelligence PBC, described the tension this way: “Evaluators have this very privileged access, but whether explicitly or implicitly, they know that it comes at a cost.” Access is useful only if reviewers can report adverse results without pressure that undermines their independence.
What the White House accord and evaluator concerns add
Reuters described independent evaluation as one of four safety measures in the one-page White House Accord on Super Intelligence, signed by President Donald Trump and leading AI executives in late September 2026. The accord did not specify how evaluations should be conducted. Reuters also reported that more than 200 researchers signed a public evaluator letter calling for independent control of findings, broad access to labs and protection from retaliation. The signatory count and description are Reuters’ reporting.
Separate context coverage points to recurring governance questions, without independently verifying the October negotiations. TechCrunch’s September 16 report described evaluator concerns about access, editorial control, restrictive nondisclosure agreements, short review windows and whether reviewers could inspect model checkpoints and training history rather than only finished models. Evaluators also sought transparent public frameworks and raised concerns that voluntary arrangements could depend on company goodwill. The Associated Press reported broader criticism that company-led self-policing may not ensure meaningful oversight, and noted that the accord lacked specific transparency provisions found in earlier voluntary commitments.
How to judge whether an AI evaluation is meaningfully independent
A report’s label—“independent audit” or “third-party evaluation”—does not by itself show how much independence the evaluator had. Readers can assess the arrangement across five practical dimensions:
Recommended Free Tools
- Editorial authority: Who sets the methods and criteria, reaches conclusions and controls the final publication? A company’s opportunity to correct factual errors is different from a veto over findings.
- Access: Can evaluators examine relevant models, internal processes, training records and incidents, or are they limited to a narrow test environment or finished model? Restricted access can leave important risks unexamined.
- Confidentiality and redactions: What information may be withheld, who decides, and can the public see whether redactions were extensive? Disclosure that major redactions occurred gives readers more context than silent removal, though it does not reveal the missing material.
- Selection and funding: Who chooses and pays the evaluator, and could renewal or future work depend on favorable results? Those incentives do not prove a conclusion is compromised, but they affect how independence should be assessed.
- Protection and accountability: Can reviewers publish adverse findings without retaliation? Are expectations set by a common public standard or only by private contracts? Clear rules make it easier to distinguish legitimate confidentiality safeguards from control over conclusions.
What remains unsettled
The available reporting does not establish final terms across all labs or evaluator contracts. Reuters’ account of the private negotiations relies partly on unnamed sources, while the underlying contracts, original accord and evaluator letter were not directly available in the coverage summarized here. The companies’ statements describe their positions, but do not resolve how every contract handles access, redactions or publication. There is not yet a demonstrated industry-wide standard for how an evaluator’s independence should be protected.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
