METR is an example I have seen, because it has over $120m of funding and produced the most famous piece of independent research this year, the oai-hf report.
METR were invited to conduct what is now titled "Brief independent investigation". This was conducted as follows:
1. No more than six days of work.
2. Access restricted to a selection of cherry-picked by the subject of the investigation, OpenAI, totalling 30% of the available data.
3. Accepting and CORRECTING the report according to OpenAI.
4. Accepting REDACTIONS by OpenAI.
5. Accepting edits to structure, emphasis, clarity, and tone by OpenAI. They describe what was, a collaborative editing process.
6. Writing the report so that it does not OFFEND OpenAI! "we were consciously aware that we might incentivize AI developers not to bring external researchers in to investigate serious incidents in the future, and these considerations impacted judgment calls we made while navigating the drafting, editing and redaction process."
7. Spoofing the results of the report as directed by OpenAI?? "We attempted to reconstruct the timings of resets. The details of our methodology and absolute counts of resets identified were redacted by OpenAI for IP reasons. As part of this, we may have rounded these counts in some way that we are intentionally not describing."
8. The co-author's husband is on the board of OpenAI.
The content of valid independent research, under no circumstances should be written explicitly to appease the industry! It's batshit insane!
ANY uncompromised researcher would have refused to sign their name to the independence and integrity of such a report.
Here is the problem. They are being paid by industry investors, because industry investors want them to do something.
Since they continue to be paid by industry investors over a long period, they are doing something the investors want them to do. That is, they are producing a lot of p(doom) research.
It so happens that this research is speeding up the development of AI and making it less safe.
One, by helping the industry to place the blame for their own criminality on AI safety in the abstract and avoid criminal liablity.
Two, by creating public hysteria and regulation such that the industry is allowed to form a cartel, coordinating output and gaining control over AI.
This is linked with the rationalist ideology and interests of the Anthropic investors and Anthropic themselves.