Report says AI startups are sharing less research in public

In short: More AI startups are keeping their research private, and one startup says it stopped a project because it could not judge the risk.

What's going on

A report discussed by The New York Times says many top AI startups are publishing fewer scientific papers than they used to. This has raised concerns about transparency, meaning the public cannot easily see what these companies are working on.

The analysis behind the report found that over half of “AI unicorns” had never played a major role in publishing a scientific paper or a preprint. A unicorn is a private startup valued at over $1 billion. A preprint is a research draft shared early, before it goes through formal review (like showing your work before the teacher grades it).

The report also included an example from an unnamed AI startup. The company said it shut down a piece of research because it could not tell whether the work was legitimate or nefarious. In plain terms, it could not decide if the work would be used for good reasons or harmful ones, so it stopped rather than take an unknown risk.

What to watch

If more AI work stays behind closed doors, outside experts will have a harder time checking results, spotting mistakes, or warning about possible misuse. Watch for whether governments push for more disclosure, and whether companies create clearer ways for independent researchers to review safety without making sensitive details public.

Source: NYTimes