OpenAI has been giving this as their reason for years and nothing has happened even though powerful models were released open source.

Turns out that misinformation is already incredibly cheap to create and the bottleneck lies elsewhere.