Surveillance without borders, weaponized cameras, and the Russian bot myth: Best infosec long reads 10/17/26

10/17/26: After spending a good deal of time this morning steeped in disturbing stories about the spread of surveillance technologies, I felt compelled to open up this week’s long reads to all visitors. These stories span China’s pervasive monitoring of its own citizens; Uganda’s use of foreign surveillance tools, including Chinese technology, amid the Trump administration’s retreat from support for human rights; the commandeering of civilian cameras during wartime; and corporate America’s growing ambitions to monitor customers.

We often encounter these kinds of stories individually, but this week’s long reads provide a serendipitous snapshot of the collective scale of surveillance and its huge ramifications for privacy, freedom, and human dignity. I think this larger picture deserves everyone’s attention, which is why this week’s full issue is available without a paid subscription.

In addition, this week's issue offers expert insight into why Ukraine's attacks on Yandex's data center sadly didn't reduce Russian misinformation bots and flags how a top machine learning researcher's monster essay is rocking the AI world, with potential ramifications for cybersecurity.

Video Cameras Are Everywhere—and Now They Are Being Hacked for War

The Wall Street Journal's Daniel Michaels, Anat Peled and Milàn Czerny, with rich graphics storytelling by Audrey Valbuena and Max Rust, explain how militaries and hackers are commandeering poorly secured civilian cameras to track troop movements, guide attacks and assess damage, with AI making this surveillance easier to conduct at scale.

Russia and Ukraine have ramped up camera-hacking in their war, aiming to spy on each other, find new targets and assess damage from their attacks. Israel, long a master of digital spycraft, hacked into Iranian traffic cameras to help forces collect intelligence used in the assassination of Iranian Supreme Leader Ali Khamenei earlier this year. Iran and its proxies have struck back by breaking into hundreds of Israeli webcams and security cameras that Israeli officials consider sensitive to national security, like the one on the chicken coop.

“They can’t look through binoculars” to see Israel from Iran, said Eliav Noori, a cybersecurity expert at Israel’s cyber directorate. “So they take advantage of cameras that are exposed to the internet and turn them into lookout points.”

That’s how Team OneFist helped Ukrainian forces find and attack a secret Russian air base in late 2022.

In 2024, Russian hackers commandeered residential webcams in Kyiv, including one installed by a condominium association to monitor the neighborhood. Ukraine’s security service, the SBU, said that Russia intended to use the hacked video feeds to find targets in the capital and “to adjust missile attacks on Ukraine,” refining their aim for subsequent strikes. The SBU said it has identified and secured, reoriented or blocked roughly 40,000 video cameras that posed potential security risks since Russia’s 2022 invasion.

Russia “is constantly trying to obtain intelligence to prepare further strikes on Ukrainian cities, including through unauthorized access to video surveillance systems,” the SBU said. 

Russia’s efforts extend beyond the war zone. Dutch intelligence agencies warned in July that “Russian state actors are systematically conducting digital espionage operations” by hacking into cameras across Europe. The operations aim to collect data on military transport routes, arms deliveries to Ukraine and the whereabouts of Ukrainian military personnel, the agencies said.

Cybersecurity consultants CloudSEK recently discovered an exposed server they suspect belongs to a Russian-speaking hacker that contained hundreds of webcam images of Ukrainian facilities involved in defense, government, energy and communications, plus assorted views of daily life. CloudSEK said many of the images came from locations near Ukraine’s borders with NATO ally Romania.

Webcams are also playing an increasingly important role in the war with Iran. An Iran-linked group called Moses Staff hacked into a Jerusalem security camera to monitor the 2022 bus stop bombings by a Palestinian militant that killed two people and injured more than 20 others, Israel’s cyber directorate said.

Where Every Tree Hides ‘an Enemy’: How I Watched China’s Paranoia Grow

The New York Times' Andrew Higgins recounts how, after returning to China after decades away, he finds extraordinary economic progress accompanied by pervasive surveillance, political paranoia, and an increasingly fearful silence.

Based in Shanghai, I often went on weekends with my wife to a park in the city’s northern Hongkou district, where groups of older Chinese gather to sing songs from their youth. Their favorite numbers are patriotic anthems but the singers, who often included a Uyghur man dressed and made up to look like Charlie Chaplin, were there of their own accord.

Their boisterous camaraderie and national pride seemed entirely genuine.

An official sign nearby, however, clouded the carefree mood: “Opposing Traitors and Guarding Against Spies Starts With Us,” it warned, with a phone number to “report espionage activity.”

The hunt for spies has escalated sharply since Xi Jinping took power in 2012 and began throttling the flow of information in and out of China, building up the Great Firewall to prevent Chinese from viewing many foreign websites.

There are seemingly no limits on what the state can see and hear, and ordinary Chinese have adapted accordingly — most noticeably with regard to the leader himself.

Mr. Xi is everywhere, on television, on billboards and on the front pages of newspapers and websites. But, throughout my stay in China, not one of the hundreds of Chinese I spoke to, mostly in private, ever uttered his name out loud. Even foreigners often avoided it, pointing at the ceiling when they wanted to refer to China’s president.

The Chinese people whom I met on my travels invariably skirted politics, except when I encountered a zealous patriot during a visit to Dazhai, a former Maoist model commune in a remote mountain area of Shanxi. “Which is better, China or America? Tell the truth!” he shouted, in between praise of China’s leadership and ridicule of the United States and Europe.

Nobody voiced open political dissent but many grumbled about the economy, particularly in places like Hainan Island where a runaway property boom has ground to a halt. A retiree from Shandong who had spent all his savings and borrowed money to buy a Hainan apartment that later lost more than a third of its value told me his much younger wife, his second, was threatening to leave him because he had fallen deeply into debt.

The Authoritarian Future Is Already Here

In a long essay reported from Uganda, the New York Times' Megan K. Stack examines how foreign surveillance technology, cooperation among autocrats, and retreating American pressure on human rights are helping authoritarian rulers tighten their grip.

Ugandans have discovered that, increasingly, there is nowhere to hide. Civic life here is defined by thickening webs of state surveillance built with foreign help. Chinese cameras and facial-recognition software monitor the streets; Russian radio frequency chips implanted in license plates have been tracking cars in real time; Israeli software plunders the private contents of phones. Technology has created an intrusive eye that never blinks — a stealthy weapon for dictators keen to stifle speech and thought.

Welcome to our menacing new world, a place where authoritarians — established and aspiring alike — enjoy tolerance from the international community; wield cheaper, quieter and far more powerful surveillance tools at home; and reach across borders to help one another hang on to power.

Africa has become a testing ground for all of this. Authoritarianism is spreading on the continent, where Freedom House’s 2026 report found that only 10 percent of people now live in countries classified as “free,” a dismal statistic outdone only by the Middle East.

China has quietly fueled this political hardening. With unparalleled trade reach, a voracious appetite for natural resources and a compelling authoritarian model, Beijing has deepened its economic and political presence across Africa. Along the way, it has exported many of the technologies that the Communist Party uses for political and social control at home. African governments have spent at least $2 billion on the Chinese artificial-intelligence-powered surveillance systems that hound dissidents from Mozambique to Zimbabwe to the Republic of Congo.

China is not the only one. Israeli spyware such as the notorious Pegasus piracy software, which can use a phone to spy on its owner, has been snapped up by rulers around the world. And companies from the United States, Canada, Britain, France, Germany and other European states have exported surveillance technologies to Africa and other regions. Such tools have sometimes been used to target opposition figures, journalists and activists.

I visited Uganda recently to explore the aftermath of President Trump’s abrupt retreat from human rights, clean government and public diplomacy. I was told, over and over, that American silence during this clampdown has made it easier for other Western countries to remain similarly quiet and has radically emboldened an increasingly ruthless Ugandan regime.

This is not to overstate past U.S. altruism. America has always chased its interests overseas — sometimes violently, often cynically, selectively calling out the abuses of disfavored regimes while ignoring repression carried out by strategic allies. But it’s also true that America stood up for dissidents and human rights defenders when it was not inconvenient and sometimes when it was. It pressured despotic governments, criticized rigged elections and imposed sanctions on officials.

Now, suddenly, Beijing’s rules prevail: Make money, extract resources, turn economic ties into leverage. Whatever you do, respect local power; don’t criticize other governments for tyranny or abuse. Mr. Trump plays by those rules, too — “trade, not aid,” his administration pithily calls it.

“Customer Surveillance” Startup Says “Every Company Will Soon Assign an AI Agent to Every Customer to Watch Their Every Move”

Futurism's Maggie Harrison Dupré reveals how Y Combinator-backed Outlit proposes using AI agents to continuously compile information about business customers from internal records and public sources, without notifying those monitored or offering an opt-out.

Outlit’s goal is to amass as much real-time information about a company’s clients as possible. According to [Josh Earle, Outlit co-founder], agents trawl internal records like call logs and sales data as well as external information like social media posts and the websites of people’s employers; they then compile that data into detailed customer profiles, often complete with “actionables” like outreach recommendations and sales guidance. Outlit argues this level of insight will help companies prevent “churn” — business jargon for lost customers — or find opportunities to upsell.

“Every company is looking to automate so much customer work with agents, but there’s a whole issue of data freshness, real-time understanding of customers,” Earle said. “What we found was that if we were able to basically monitor a customer across every internal and external signal, we could actually create a real-time understanding, and this would include monitoring all the activity under product, under [customer relationship management], under LinkedIn, under website — everything.”

Outlit, which is backed by the incubator Y Combinator, is yet to be publicly released. And as it stands, it’s working mostly with business-to-business software companies, and says it isn’t in the business of monitoring individual consumers — nor does it want to be, according to Earle.

“We probably wouldn’t be going business-to-consumer,” Earle told us. “The data would suck anyway.”

But while Outlit isn’t sending agents to follow random shoppers around the web, said Earle, the goal is to “be monitoring every external data piece from a company, and the internal as well… they need to know what’s going on with [their customers] — from every call that they’ve had, from every interaction, from every LinkedIn.”

The resulting product, as described by Earle, is basically a detailed intelligence dossier.

“It’s one page, and it’s always updating,” said the founder, “and it’s all of everything we can find on them at all times.”

It’s true that companies already store a great deal of data about their clients, scouring the web for information about a given customer and its workforce; Outlit argues that there’s too much valuable data out there for individual human workers to effectively gather, let alone make sense of quickly enough for it to be relevant. But as with other AI-powered surveillance technologies, the deeply-automated process adds an extraordinary layer of scale and perpetuity to data collection and consolidation.

Agents, after all, never need to sleep.

The Russian Bots Went Silent. Or Did They?

Famed misinformation researcher Renée DiResta, an Associate Research Professor at the McCourt School of Public Policy at Georgetown, explains why a dramatic decline in Yandex web-crawler traffic does not establish that Ukrainian strikes silenced Russian propaganda accounts, while shrinking platform data access makes such claims harder to investigate.

Between Oct. 8 and 11, Ukrainian drones struck three Yandex data centers in Russia, in Sasovo, Kaluga, and Vladimir. The attacks disrupted Yandex services and affected Russian businesses that depended on its infrastructure. Reuters reported the strikes, and the Kremlin acknowledged the resulting service problems.

Cloudflare's measurements documented the disruption in real time. Following the first strike, traffic from Yandex Cloud's network (AS200350) fell approximately 33 percent below normal beginning at 22:30 UTC on Oct. 7. Traffic from Yandex's main network (AS13238) also declined, falling roughly 25 percent below its earlier baseline on Oct. 8 and 35 percent on Oct. 9–10, according to one analysis of Cloudflare data.

By 2 a.m. UTC on Oct. 11, there was evidence of a dramatic collapse: automated HTTP requests from AS13238 had fallen to just 0.3 percent of the chart's recent maximum, compared with 88.2 percent at the same hour a week earlier. This is a decline of more than 99 percent relative to the comparable hour.

That's an enormous drop in automated traffic; the surprising chart was shared online by members of the public, along with visuals of a burning data center. Yet the broader viral story hinges on what, exactly, Cloudflare’s chart means by a “bot”—and whether it's the same thing people mean when they complain about “Russian bots” on social media.

“Bot” is shorthand for an automated entity that performs tasks online. This can mean many different things depending on who is using the term. Search engines such as Google and Yandex deploy web crawlers—often called bots—to visit websites and index their content; Google sends Googlebot to scan news websites for example, so that their articles can appear in search results. Cloudflare tracks this kind of traffic.

Social media bots, meanwhile, are accounts that automatically post, reply, or amplify content. On social media, the term “bot” is used much more loosely. Some accounts are automated, and automatically post or amplify content. Some combine human operators with automated activities, and “bot” stands in for “troll.” When people complain about "Russian bots" in their replies, they're often using the term to describe accounts they suspect are fake or inauthentic, regardless of whether they're actually automated. And sometimes they're just insulting someone they disagree with.

Neural Networks Are Already Clean

Machine learning researcher Nick Cammarata of Goodfire released a 26,000-word opus this week that has, as The Atlantic put it, set the AI world on fire as the "most important essay" in the history of AI research, with seemingly powerful implications for cybersecurity, including possibly deeper investigation of AI failures, with causal evidence supplementing behavioral testing.

Behind doubts about small errors in neurons is a fear that safety will come down to some activation we missed, the place where deception hides. The risk I worry about is different: like being in a well-organized city and not knowing how to think about it, because we aren’t native to it. Every error discussed in this essay was something hiding in plain sight that we didn’t know how to read, or that our tools read incorrectly. None was a hidden activation we missed. Deception might hide from someone reading a model’s outputs, but it has no reason to hide from the model’s own circuitry. It needs to be accessible to its readers, and if we become a reader, it should be accessible to us too. My fear is that deception will be right there, in plain sight, and we won’t know how to read it.

Over time, my approach to individual mechanisms has become less focused on this kind of proof, and more holistic. Instead of trying to nail every detail of one neuron, I look around it: what flows into it, what it feeds, how attention moves information to and from it, and what it looks like zoomed in and zoomed out. When I worked on automated interpretability, a neuron firing a few tokens away from its main topic seemed like a big problem. But over time it was clear that attention moves things to reliable places, which then have neurons that activate near the “appropriate token” rather than on it. That’s how understanding has felt to me: less like clearing a bar for each piece and more like a fog-of-war map filling in at several levels at once. Attribution graphs help explain neurons, neurons help explain attention, and attention explains neurons in turn.

Neural networks rarely get stuck during training, since in a high dimensional space there’s almost always a way downhill. In my experience, clearing the fog of war works the same way. There’s always another experiment to run, another explanation to narrow, another attribution graph to check, another neuron to watch grow over training.

This is why I’ve focused on whether there’s any level of abstraction where things break down, where you stop being able to learn anything and have to take the network apart into something else. I haven’t found one. And when each new motif clears fog in many places at once, perfecting a single neuron probably isn’t the best use of effort. There’s a kind of faith in it: you note something strange, and over time it usually fills in. It isn’t faith all the way down, since almost any explanation can eventually be tested. Just faith that something strange is a good pointer. The V in the car and cat neuron isn’t hard to see – it took me about five seconds, and most people I’ve shown the image grid to and say “what’s the shared pattern here” see it too – but it took believing that the neuron had found something real. For years I’ve heard this neuron come up in almost every discussion of polysemanticity. It was almost always described in reference to neural networks being almost invalid objects of direct natural science. That’s what I was saying too, until I looked more closely. It didn’t take a lot of faith, but it took a little.

The little faith is important because the mistakes aren’t symmetric. If you assume a network is clean and you’re wrong, a messy neuron is a lead: you investigate it, and over time normal science sharpens your explanation. If you assume it’s packed, not worth studying directly, and you’re wrong, a messy neuron is the end, because you stop studying it. I made that second assumption for years, and it explained away the evidence that would have overturned it. I could have improved my tools, watched the mixtures keep dissolving, and built up a prior that the neurons could be trusted. Instead, I spent most of a decade figuring out how to look at a model rather than looking at it.