OpenAI busts influence ops that planted fake stories in real media
The day's AI news runs through one seam: the work is showing up in places nobody planned for — inside real newsrooms, across the whole night sky, and in the M&A column.
OpenAI has banned two state-backed influence operations that used ChatGPT to plant fabricated stories inside legitimate news outlets — and rated the Russian one the most disruptive it has seen in two and a half years. In a report dated October 8, OpenAI detailed "Dark Clark," run from Russia across Latin America, which ran a think tank through a fictitious persona and pushed fabricated audio and documents that triggered fact-checks and official denials in Ecuador and Peru; OpenAI scored it category 5 out of 6 on the Breakout Scale — the first category 5 in the scale's two-and-a-half-year reporting history, precisely because politicians reacted to the content. The Iranian operation, "Bogus Bylines," used seven fake journalist personas to place nearly 100 articles about the US-Iran conflict in online outlets worldwide; the models handled drafting, editorial pitches, and multilingual adaptation, while the operation's generated social-media comments sank without a trace. The tell is the target: both campaigns skipped the usual bot-farm playbook and aimed for the bylines instead — planting stories where editors and readers already extend trust is a harder problem for fact-checkers than another flood of inauthentic posts, and it moves the cleanup work from platforms to publishers.
Anthropic says an astrophysicist used Claude Science to produce the first complete map of the sky in ultraviolet light. Johns Hopkins's Brice Ménard lays out the project in an Anthropic research post: Claude Science coordinated AI agents that downloaded data from multiple space missions, calibrated and merged it, and filled the gaps with inpainting — in tests, predictions averaged about ten percent deviation from actual measurements. A complete UV map didn't exist until now because the ozone layer blocks ultraviolet light from the ground, and NASA's GALEX mission covered only about two-thirds of the sky while skipping bright star-forming regions. Anthropic frames it as teaching material; the wider signal is that projects researchers had shelved as too tedious to finish are now within reach of agent orchestration.
Venture-backed AI companies have completed 195 acquisitions so far this year — 14% more than all of 2025 — with OpenAI leading the spree. Crunchbase's data, running through September 29, shows the buyer pool barely grew (up 2%); the same companies are simply buying harder. OpenAI has completed 20 AI-related acquisitions over three years, 10 of them in 2026, ahead of Anthropic and legal-AI startup Legora with five each; Harvey follows with four, Sierra and Cursor with three apiece. Only 12 of the 195 deals disclosed a price — the largest was Nscale's $1.65 billion purchase of distributed-computing firm Anyscale, then Cyera's $1 billion for Oasis Security, Anthropic's $400 million for Coefficient Bio, and OpenAI's $300 million for Glass Imaging. With prices hidden in most deals, the honest read is volume over value: buying instead of building has become the default operating strategy for well-funded labs and vertical players alike.
What to watch: whether more of these operations surface as regulators start asking labs for threat-report access — and whether OpenAI's Breakout Scale score prompts other labs to adopt a shared rating.
If influence campaigns are moving from fake accounts into real publications, who should be checking the byline — platforms, publishers, or the models themselves? Tell us in the comments.
