An AI-assisted influence operation got almost 100 articles published or syndicated under seven fake journalist identities, according to OpenAI's October 8 report.
The operators used ChatGPT to edit English-language drafts against individual publications' submission rules, then write pitch emails to their editors. OpenAI says it traced the bylines across roughly a dozen outlets covering international affairs. Some outlets then promoted the articles through their own social accounts.
That gave the operation something a standalone bot account has to struggle for: distribution through an existing publication. The articles appeared under journalist bylines on existing news outlets, giving them a route to those outlets' audiences. The identity supplying it was the part that had been fabricated.
The same operation also generated social media replies. OpenAI found little engagement on the replies it could identify, and says some praise beneath posts promoting the planted articles came from the operation itself. An article and its approving comments could therefore come from the same source while looking like separate signals of credibility.
This was a geopolitical campaign, not a documented crypto promotion. The report doesn't establish how many readers believed it. But it gives anyone following news through a feed a specific problem to recognize: publication and apparent audience endorsement can both be arranged by the people behind the story. Checking whether the underlying claim has an independent source matters more than counting the places it appears.
Source: https://openai.com/index/disrupting-ai-enabled-false-front-operations/
Image: AI-generated editorial illustration, not an actual newsroom.
The operators used ChatGPT to edit English-language drafts against individual publications' submission rules, then write pitch emails to their editors. OpenAI says it traced the bylines across roughly a dozen outlets covering international affairs. Some outlets then promoted the articles through their own social accounts.
That gave the operation something a standalone bot account has to struggle for: distribution through an existing publication. The articles appeared under journalist bylines on existing news outlets, giving them a route to those outlets' audiences. The identity supplying it was the part that had been fabricated.
The same operation also generated social media replies. OpenAI found little engagement on the replies it could identify, and says some praise beneath posts promoting the planted articles came from the operation itself. An article and its approving comments could therefore come from the same source while looking like separate signals of credibility.
This was a geopolitical campaign, not a documented crypto promotion. The report doesn't establish how many readers believed it. But it gives anyone following news through a feed a specific problem to recognize: publication and apparent audience endorsement can both be arranged by the people behind the story. Checking whether the underlying claim has an independent source matters more than counting the places it appears.
Source: https://openai.com/index/disrupting-ai-enabled-false-front-operations/
Image: AI-generated editorial illustration, not an actual newsroom.