OpenAI Bans Accounts Behind "No Bell" Influence Operation

OpenAI has banned a set of accounts tied to a previously unreported influence operation, disclosed under the name "No Bell," that used AI tools to generate criticism of the United States and its allies aimed at audiences across Africa.

OpenAI said the operation was "likely Russia-origin" and marked the first public reporting of the network, which had not previously been documented by the company or, it appears, by outside researchers. The disclosure adds a new named campaign to the growing list of state-linked influence operations that have incorporated commercial AI models into their workflows.

What do we know about "No Bell"?

According to OpenAI's disclosure, the operation consisted of accounts on its platform that used AI to produce content criticizing the US and its allies. The material targeted readers across the African continent. OpenAI assessed the operation as likely originating from Russia.

The company described the network as "previously unreported," meaning "No Bell" had not appeared in its earlier threat-intelligence publications before this takedown. OpenAI's investigators assigned the operation its name as part of standard practice for tracking coordinated networks.

The scale of the operation — follower counts, posting volume, platform reach — was not specified in the initial disclosure.

Why does this matter?

The case matters for two reasons.

First, it extends a documented pattern: influence operators linked to Russia, China, Iran and other states have repeatedly turned to consumer AI services to scale content production, refine translations and simulate engagement. Each new named operation shows commercial model providers are now positioned as an early detection layer for covert propaganda, catching networks before or alongside traditional social-platform takedowns.

Second, the African target audience fits a broader strategic picture. Russia has invested heavily in media and information operations across Africa in recent years, framing the US and Western partners negatively while presenting Moscow as a partner. AI-generated content lowers the cost of producing polished, localized material in many languages at once — which is precisely the capability influence operations want.

How did OpenAI find it?

OpenAI's safety and investigations teams routinely hunt for clusters of accounts that violate its policies against covert influence and deceptive use. When they identify such clusters, they ban the accounts and publish reports describing the operation's goals, origin assessment and infrastructure.

The "No Bell" disclosure follows that template: a named operation, an attribution judgment ("likely Russia-origin"), a description of intent — criticism of the US and its allies — and a statement of target audience across Africa.

What comes next?

OpenAI's disclosure puts "No Bell" on the record for other platforms, researchers and governments, who can now look for matching infrastructure and content signatures outside OpenAI's own ecosystem. Expect follow-on reporting from threat-intelligence firms and, potentially, additional takedowns across social platforms as the operation's footprint is traced.