When your TV watches you

Gamers Nexus reported weeks ago that some LG televisions and monitors are intruding on customers’ lives in unsetting ways, from inserting recurring ads into monitors to possessing the capacity to record private conversations you have in your home. The story got more traction in recent days, with several “Can they really do that?’ posts.

In another development last week, mathematician Tristan Buckmaster wrote that OpenAI announced its AI had solved a longstanding problem—one the company knew Buckmaster and Levent Alpöge had been working on. The pair had been using several LLMs, including one from OpenAI, as a tool to assist them. Buckmaster made no direct accusations, but the letter raises implicit questions about whether someone at Open AI accessed the mathematicians’ work intentionally or inadvertently in order to hijack the humans’ discovery for the company.

And just this week, 404 Media posted a story about private human contractors involved with what’s known as “Project Lily.” They’ve been hired to go through users ChatGPT chats—conversations which often include highly personal data—without users’ knowledge.

I’m not a tech expert, so I’m going to come at how all these are related a little sideways. My intent is to stay in my lane and address what the issues raised by these examples mean in the current political climate. I’ll consider some historical examples of authoritarian surveillance, what it means to have a government that is seeking to oppress millions within a country’s borders alongside private companies looking to harvest personal information for profit, and how corporate overlords and aspiring dictators wind up collaborating in a police state.

In a post last year, I mentioned a complaint made in 1937 by Chief of German Police Heinrich Himmler. Himmler lamented how insufficient his surveillance tools were in the face of so many enemies of the state.

“What does it mean, putting a man under surveillance?” he wrote. “For that I need at least three officials per day—there are 24 hours in a day—these three officials need two cars, because if the fellow is at all clever, he’ll leap from one streetcar to another… So it can’t be done without five officials.”

Every police state is limited by its ability to track everyone it wants to. Different authoritarian governments respond to that challenge in different ways. In East Germany during the Cold War, the government spying apparatus employed a quarter-million employees, the majority of whom were informants—estimated to be 1 in 40 people in the country. Including part-time and unpaid spies, estimates of the overall number of informants have gone as high as one in every 6.5 citizens living in East Germany at the time.

While the U.S. never reached those levels in the same period, the FBI’s COINTELPRO operation worked to spy on, “misdirect, discredit, disrupt, and otherwise neutralize” those it saw as enemies. From the Ku Klux Klan to feminists and leftist students, the FBI conducted unethical and illegal surveillance on countless groups, but particularly targeted Black activists and community leaders.

When COINTELPRO stood revealed in the 1970s, backlash followed, and guardrails were installed. In some cases, those guardrails have since been dismantled. In other cases, technology has simply vaulted beyond the legislation from that era. I won’t go into all the failed privacy battles and lapsed protections here. The important thing to understand is that new tech has made surveillance possible on a level that Himmler never dreamed of.

The state-corporate nexus

In the twenty-first century, China has built a staggering surveillance system pioneered for use against the Uyghur population in the autonomous region of Xinjiang. That system has become a pervasive and invasive collaboration of companies with party leadership, and has provided tech and/or served as a model for repressive governments elsewhere. Among other countries, facial-recognition technology has been exported for use in Venezuela and Zimbabwe.

The U.S. and Europe have taken action against some of these Chinese companies. But simultaneously, tech developments by companies based in the U.S. represent their own category of rising threat to every person on the planet. But maybe not for the reasons you’re hearing about most.

You may recall that for taking a stance against its technology being used for autonomous killing machines and mass domestic surveillance, Anthropic was blackballed by the government. Unfortunately, other AI companies stepped in where Anthropic wouldn’t go in that moment.

Yet Anthropic has played its own role in problematic police-state issues—though perhaps one it had not intended to. The company reported just days ago that its Claude AI models had been used overseas, including in China, for activities ranging from “weapons development and cyber operations to surveillance and fraud.”

Not every surveillance company is primarily an AI company, of course. Just this week, WBUR in Boston broke the story that “three days into a pilot program with the Boston Police Department, city officials realized Flock Safety had been distributing data it promised not to share with external law enforcement agencies, according to the city’s 2025 Surveillance Technology Report.” The sharing of data had been specifically prohibited in the city’s trial contract with the company, yet other agencies had search the data from the new Boston cameras repeatedly.

An Electronic Frontier Foundation data project also published this week discovered that despite official requirements for a lawful reason to search license plates when searching Flock camera databases, officers frequently typed in phrases like “LMAO,” “IDK,” and “Hehe.”

Meanwhile, Elon Musk’s Grok/xAI is facing new accusations that its artificial intelligence has produced child sexual abuse material based on photos of actual crimes. And last week, Ars Technica reported that Meta ads pushed apps nudifying real teens. The apps allowed clients digitally alter images of children using AI, and Ars Technica found that the photos in the ads were manipulated versions of identifiable photos of real teens findable online.

These are all different kinds of surveillance for profit. Authoritarian governments are more interested in controlling you than protecting you. Companies are less invested in that agenda directly, but in many cases, stand to gain from it.

So much data is now collected on every one of us; yet it’s imperative to set limits where we still can. In the case of LG TVs, the televisions have the capability of monitoring other devices in the household and recording conversations in the room. While gathering this information for their own purposes, companies effectively surveil their clients.

Surveillance becomes valuable in several ways. By watching you more, companies can come to understand how to better manipulate you and make you more dependent on them. In theory, this information could be used to better serve your needs. But what we’ve seen a lot in the last decade is that through the process of what’s come to be called enshittification, the more dependent you are on the company’s product, the less the quality of their product will matter to them. They’ve already captured you.

So in truth, you can just wind up in thrall in a different way than you are with surveillance done by the government. In the latest example, the AI-ification of life, externalization of your own skills and knowledge has real risks. But when surveillance puts you at the mercy of a police state, that’s another level of danger altogether.

What can you do?

On the national and international level, surveillance tech needs to be regulated. But what if the governments involved are a threat to their own constituents? Who would you trust in the federal government today to regulate artificial intelligence?

Senator Bernie Sanders of Vermont is arguing for greater control by the government, but to do so, he partnered this week with Steve Bannon, a proven demagogue and threat to civil rights across the board in the United States. Working across the aisle is one thing. Working with someone who tried to sabotage the country is something else altogether.

That said, the tech oligarchs seem to be using a potential AI safety pause mostly to block out their overseas rivals. It seems hardly anyone in power is approaching the issue in good faith.

There’s likely not time to stake the surveillance issue out in a thorough and comprehensive way before the November elections. But it should be on national politicians’ radar in 2028 as part of an agenda of revitalizing democracy (a theme the House Democratic super PAC leadership has been hostile about spotlighting for the midterms).

On the household level, there are simpler things you can do. Turn off the microphones on your TV if you don’t want it to have the option of eavesdropping on you. Tom’s Guide has a good post from last year on how to do that. (But be aware that you’ll also disable the voice features on your remote.)

Be aware of how often you are outsourcing your skills and ability to do tasks that AI cannot do as well as you or which will degrade your own abilities over time in ways that matter (see, for instance, the effects of using AI on students and doctors).

However you feel about new tech as a concept, always ask yourself what the risks are to incorporating it on an institutional level. Should it really be in your schools or universities? At what age? What happens to the data generated by those students?

Should it be in your workplace? What liability issues does that raise, if the AI companies can access that data and turn it over to other people? How can we protect people’s private sessions with agents from being sold or seen by others without the user’s consent?

Remember that once doors are opened on surveillance, they’re very hard to close. In addition, the leaders of the most prominent AI companies in the U.S. are for the most part, bad actors who are actively destabilizing democracy—and, in many cases, the planet. Their leaders are not trustworthy on matters of surveillance.

One of the tremendously disappointing, though not surprising, developments of the last decade has been the ways that companies have utterly failed to stand up to the dangers represented by authoritarian states. And with cutting-edge technology, aspiring dictators won’t need one out of 6.5 people to become informers, as happened in East Germany during the Cold War period. They won’t need a massive police force of dedicated secret police to trail suspects, as Himmler longed for in Germany (although they clearly have one with ICE). They just need people to want more new toys badly enough to ignore they’re being watched for long enough to establish a seamless network that will be too entrenched to undo.

Donald Trump said yesterday that the only regulation AI needs is a smart president with a High IQ. But we already know that the worst surveillance technology has been used in damaging ways to kill people even by smart presidents who imagined themselves to be good actors. And if there’s anything that Donald Trump’s political career has taught us, it’s that in the hands of openly corrupt leaders and amoral oligarchs who bend the knee to authoritarians, the risks of giving these people any new power are exponentially higher.

I’m not trying to terrify you, but instead to motivate you. Harder battles than this one have been won before. You don’t have to believe even a single apocalyptic sci-fi scenario about sentient AIs (I don’t) to understand that it’s the oligarchs developing tools with surveillance capabilities and the corrupt officials in positions of power—those who are eager to exert control over the public—who are the real threat.

Your paid subscriptions support my work.