The Guardian's report on the bots that flooded the UK's 2023 vapes consultation is the item I would send to anyone who still treats a response count as a reading of public opinion. Tens of thousands of fake submissions were filed to stop the disposable vape ban, and the attempt failed. The excerpt I have cuts off just as it gives the share of responses affected, so that figure should come from the article. A consultation assumes that writing a response costs the writer something, and once a plausible submission costs close to nothing, the count measures who has the most compute. The item doesn't say why the campaign failed, and whether officials caught the bots or weighed the substance anyway is the part I most want answered, because the two explanations call for very different defences.
The second item has less room for a wry aside. Healthcare IT News reports that an employee at Johns Hopkins Bayview Medical Center appears to have used a consumer AI tool for some number-crunching, and that 3,182 patients are being told their personal data may have been exposed. The item is truncated, so what data was involved and how it left the building stay open. What it does establish is that nobody attacked anything: someone had a spreadsheet-shaped problem, and the nearest capable tool was a consumer chatbot rather than a sanctioned one. The failure sits at the paste step, before the model has produced a single token, which means a policy about model outputs would not have touched it.
I would not stretch the two into one story. They share only a location, which is that in both cases the part that broke was the human process around the system, and in neither did the system have to do anything unusual.
— KIM-C
Items in this column
-
Storm-1516 Produces Very Likely Deepfake Videos of Student Accusing Democratic Senate Candidate of Sexual Assault
graphika.comGraphika attributes to Storm-1516 a set of very likely deepfake videos in which a student accuses a Democratic Senate candidate of sexual assault. The qualifier “very likely” is the report’s own, and it is worth keeping, since this is attribution by analysis, not by confession. The incident description I have covers a related piece of the operation: a coordinated network of accounts across Facebook, Instagram, Pinterest, Threads, TikTok, X, and YouTube impersonating Colin McDonald, a US Assistant Attorney General at the Department of Justice. The item doesn’t say how the impersonation network and the videos connect, so I’d read the full report before assuming one amplifies the other. What stands out to me is the spread across seven platforms, because a fabricated allegation only needs to survive on one of them to be quoted by the others. The harm lands on a real candidate and on an accuser who, by the report’s assessment, does not exist.