OpenAI banned a cluster of Russia-linked accounts that used its models to invent a nonexistent Israeli think tank.
The banned accounts generated content for a fabricated Israel-based research outfit and used it to distribute a "sovereignty index" - a ranking system that praised Russia and criticized Western governments. OpenAI says the takedown falls under its policy against using its models for covert influence operations. It did not disclose how many accounts were involved, how far the content spread, or which platforms hosted the material.
AI tools make this kind of operation cheap to run: a single model can produce a think tank's entire back catalog - reports, a mission statement, a supposedly independent index - without a newsroom or research staff behind it. That collapses the cost of manufacturing fake authority, which is the resource propaganda operations have always needed most.
OpenAI caught this one before it built a public following, but the company can only police its own product, not what a banned operator tries on the next platform.