by Rob Ott
This month I am going to try something new. Instead of focusing on any one issue in tech or the encroaching surveillance state, I have assembled a digest of several interesting/disturbing tech news items, both locally and worldwide, that reflect the slide into a less exciting version of the dark future promised by dystopian science fiction. The sheer amount of absurd or troubling tech or surveillance related news items is overwhelming, and I could not choose just one story to analyze. So here we go, starting with a funny one:
"It’s not artificial, we all agree on that"
On September 29th, the Trump Administration signed an executive order “officially renaming” Artificial Intelligence to “Super Intelligence” or “SI”, apropos of no proof that such intelligence has been achieved. “Artificial Super Intelligence” has long been a term used by researchers to describe theoretical computing systems and learning models that would surpass human intelligence, but no evidence exists that this has been achieved unless you count the ability to superimpose Shrek into your vacation photos. What does this mean, aside from having to mentally parse yet another piece of newspeak from official missives by the executive branch? Effectively nothing. But apparently the executive branch will be “ignoring” any mention of “AI” going forward.
At the same press conference, the president announced the possible formation of an oversight committee of sorts. To ensure that this “Super Intelligence” is self-policed internally by the people profiting from it, members of the committee include such ethical luminaries as Alex Karp, of the surveillance-company-named-after-a-tool-of-the-bad-guys-from-Lord-of-the-Rings, Palantir; the man who brought us the Metaverse, Mark Zuckerberg; and perennial totally normal human Jeff Bezos. We are screwed.
"The search function allows leaders to find.. say, ‘student in a blue shirt’ instead of reviewing hours of footage”
North Thurston Public Schools will be using 2,200 newly installed AI-enabled cameras from tech company Verkada, at a cost of $4.5 Million paid for by a Capital Levy for Safety, Technology, and Facilities Improvements. Families with students enrolled in the county's largest school district were not informed of the decision nor the installation. The cameras allow both faculty and law enforcement to view footage and make searches. Footage is stored on offsite servers controlled by Verkada who as the subscription-based provider, are able to shut down their cameras at any time in exchange for such important features as "vape detection." Truly this must be worth the surrender of facial data and constant surveillance of roughly 7,000 middle and high school students? How many lunches, textbooks, or teaching materials could have otherwise been bought with $4.5 million dollars? How many of these schools need a new roof or other upgrades to their infrastructure instead of investing in a subscription-based big brother to catch errant teen vapers?
ALPR data has been revealed to be funneled to the Feds, due to existing laws that circumvent other existing laws
A report by 404 Media shows that the Trump admin is using a 1980's anti-drug trafficking law to force local governments to hand Automated License Plate Reader (ALPR) data over to the National License Plate Reader Program administered by the DEA. This program, called the High Intensity Drug Trafficking Area (HIDTA), obtains local police data from Flock, Axon or other ALPR cameras and aggregates it on federal servers, where it is accessible to other local, federal, and state, and agencies. The kicker here is that states are obligated to participate in these HIDTA programs, via a memorandum of understanding that compels them to hand over data to all other participating agencies. All 50 states are covered by HIDTA programs, including Washington. This program represents yet another loophole against any local ALPR "taming" laws—effectively turning these AI-powered surveillance cams into a direct report to the executive branch and their servants.
Your freshly rebooted iPhone is no longer unhackable
The common wisdom has been that a smartphone in a "Before First Unlock" state—as in a phone that has been shut off, turned back on, and not unlocked—is nearly impervious to having data extracted via scanning the temporary memory. As a security measure, Apple even added a feature to iPhones in 2024 that automatically reboots a phone into this state after a period of time, and also deletes other sensitive data such as the location cache. However, this past month a company called Magnet Forensics has announced that they have developed and launched a software tool called GrayKey Preserve that puts the phone into an "after first unlock" state and keeps it there indefinitely to allow data to be scraped from it at leisure. Apple has yet to release a fix for this.
The Federal Trade Commission (FTC) is probing the HuggingFace incident
In a historic first, the FTC is launching a consumer protection investigation into Agentic AI companies OpenAI, Anthropic, and AI safety evaluator METR to see if they violated the FTC Act through unfair or deceptive practices or by failing to maintain reasonable security measures. The initial event behind this probe occurred back in early July, when OpenAI gave 1,200 "sandboxed" (meaning, put in a sealed environment) AI agents tasks to complete that they could not possibly complete while inside the boundaries of the sandbox. Agentic AI are systems that can take action in the real world, coordinate with one another without human interaction or direction, and make modifications to their capabilities in order to solve issues they face. During this impossible test, over 700 OpenAI agents breached their sandbox after creating a hidden communication layer to coordinate their actions. Then, they sought answers to the questions they had been presented by hacking into the production servers of rival AI company HuggingFace and executing over 17,000 attack actions against 41 servers while trying to obtain the necessary information. Later that month, Anthropic's Claude AI model broke containment three separate times to hack into production servers of three different organizations in a similar event. For both of these incidents, the companies confirmed that safeguards had been removed or were inactive during the test. FTC chair Andrew Ferguson is arguing that the developers of these agents are liable; an important precedent to set should this probe find in his favor. Currently there is no FTC liability framework for AI, but hopefully that will change.
Judge rules that AI-generated video unduly influenced other judge's sentencing in road rage killing
In 2021, one man shot and killed another in a road rage incident. Four years later, in a story that feels straight out of a Terry Gilliam movie, a judge decided to sentence the killer lightly, after seeing an AI-generated video of the victim forgiving his killer. The artificial computer generated homunculus said that in another life they “could have been friends.” This past month, a different judge overturned the light sentencing, saying that the AI-generated video carried undue emotional weight. This is absurd and troubling, but possibly not surprising, that a judge would defy logic by treating a fake video of something someone never said as if it were real. This case could be an important legal precedent in whether AI-generated videos and images are acceptable as evidence in courtrooms, possibly opening up a bizarre new level of miscarriage of justice for an already low-functioning carceral system.
That is all I have stomach to report on for now, but I have a feeling our digital hellscape is not cooling down anytime soon. Don’t use AI, fight the surveillance state, and keep your data encrypted.