The Corporate Prosecutor
When OpenAI announces that it has disrupted a Russian, Chinese, Iranian or Israeli influence operation, the public receives a tidy fable: the company spotted abuse, shut it down, and warned the world.
The omitted detail is that one company performs every role in the case. It owns the platform, defines what counts as suspicious, investigates the users, clusters accounts into a network, attributes political intent, cuts access, and publishes the verdict. Custodian of the evidence, investigator, expert witness and prosecutor share an office.
Take the May 2024 report, which bundled two Russian operations, a Chinese one and an Iranian one with a network run by STOIC, a commercial firm in Tel Aviv. The reader got the operation's name, samples of its output and a description of its behaviour. The chain linking specific accounts to a specific company stayed inside. So did everything else a finding of that weight would normally require: the logs, the full sample, the clustering criteria, the false-positive rate, the alternative hypotheses, the internal disagreements.
A platform is entitled to police its terms of service. "This account used our model to draft text" is an observable fact. "We uncovered a state-linked influence operation" is a claim about intent, coordination and political effect, and it arrives with the evidence withheld.
Intelligence agencies are bound, at least on paper, by analytic standards that require stated confidence levels and explicit alternatives. The EU's Digital Services Act gives vetted researchers a legal route into platform data. Corporate threat reporting has neither. Its conclusions are not auditable, only quotable.
Generative models make phishing, malicious code and synthetic persuasion cheaper to produce. The same industry then sells the filters, monitoring and attribution needed to contain them. No conspiracy is required. Expand the market for dangerous capabilities, then monetize the demand for their containment.
A newly unsealed brief in the New York Times case quotes Nick Turley, OpenAI's head of ChatGPT, writing internally that publishers face an "existential threat" from products that are "largely substitutive."
Microsoft's own measurements put Copilot's click-through to Times sites 87 to 93 percent below ordinary Bing search.
In court, OpenAI is a defendant whose internal documents get unsealed. In its threat reports, it is a prosecutor whose evidence never is. The company that decides who is manipulating information also replaces the sources that information came from. It moderates the adversary and supplants the witness.
That is the real subject behind the hashtag. A state that accepts corporate attribution as intelligence hands a private firm the function of naming the enemy. Whoever attributes, designates. Sovereignty over information does not end when a foreign power runs a bot network. It ends when a vendor's blog post becomes the finding of fact.
#OpenAI #ArtificialIntelligence #TechPower #DigitalSovereignty
📱 American Оbserver - Stay up to date on all important events🇺🇸