Kremlin fears that cameras are "working" for the enemy

Russian security services have turned off parts of a special surveillance system protecting President Vladimir Putin and his closest aides after the assassination of Ayatollah Ali Khamenei in Tehran, the Financial Times reported, citing sources familiar with the matter. The move came amid fears that Russia's vast surveillance apparatus had turned from a tool of control into a weak point that could be exploited by enemies.

The system, which operates separately from the nearly 300.000 cameras monitoring Moscow's citizens, was only turned back on after engineers thoroughly checked it in an attempt to completely isolate it from the internet, one of the sources told the FT.

Russian services took extraordinary precautions after learning that Israeli intelligence, by analyzing vast amounts of Iranian traffic camera footage, had pinpointed the location and time of a meeting between Khamenei and his closest associates on February 28. Several senior security officials were killed at that meeting, in the first blow of the US-Israeli war against the Islamic Republic.

The killing was a dramatic demonstration of the use of artificial intelligence to process millions of hours of video footage, collected by thousands of cameras, to find and monitor targets.

Last week, the director of Russia's Federal Security Service (FSB), Alexander Bortnikov, warned regional security chiefs that Russia's vast surveillance apparatus had become a weak point that enemies could exploit, according to Russian media.

"The recent assassination of senior Iranian officials by the US-Israeli alliance is a clear warning sign," Bortnikov said, adding that the victims' locations were partially determined thanks to hidden software accesses in Tehran's video surveillance systems.

There have been serious concerns in Moscow about Putin's personal security in the past, primarily because Ukrainian intelligence services have hacked into traffic cameras in Russia and used mobile phone location data to plot assassinations of senior Russian military officials in the heart of Moscow.

Despite Russian precautions, an independent Ukrainian hacker told the FT that cameras in Moscow, and even around the Kremlin, “are still operating and are being hacked regularly.” He declined to say whether Ukraine had the capacity to analyze them on a large scale.

Governments have long known that experienced hackers or spies, such as Israel's elite electronic reconnaissance units, can easily penetrate security cameras. However, advances in artificial intelligence in recent years have enabled them to precisely identify specific behaviors and patterns in vast sets of visual data, the FT writes, noting that Israeli intelligence used these very methods to locate Khamenei and his closest associates from Iranian camera footage, determine the time of their meeting and carry out the coup.

Israeli intelligence used AI to map Tehran's complex geography, recognize patterns of behavior among senior officials' bodyguards, and efficiently isolate targets from millions of hours of footage from thousands of cameras. They cross-referenced this information with other intelligence channels, including human sources, the FT reports.

The visual capabilities of artificial intelligence have advanced rapidly since 2023 and are now significantly more sophisticated than earlier machine learning algorithms, which were used for facial recognition, weapon detection, or vehicle tracking by make and license plate.

Unlike older tools, which were limited to a few dozen predefined searches, the new tools allow for an almost unlimited range of queries, as videos can be searched by language.

This allows intelligence officers to search through vast streams of video footage using simple search terms, such as: two men handing each other a bag; a person who has changed their appearance or changed clothes multiple times during the day; or a vehicle that has recently been repainted, or that has passed the same location multiple times in a short period of time.

“This is the holy grail of surveillance,” one European official whose country uses the technology in its cities told the FT. “We can look for behavior, not objects – it has created a whole world of new possibilities.”

New technology is effectively turning billions of dollars spent on video surveillance systems, especially traffic cameras, into a keyhole through which adversaries can search vast cities and secured facilities to extract patterns and secrets on an industrial scale.

The idea that a country's own surveillance system could be so effectively turned against it by its enemies has alarmed counterintelligence agencies around the world.

Their first reaction was to try to plug the weak points in the camera systems themselves, an extremely difficult task considering that most surveillance networks are made up of a complex mosaic of cameras of different generations.

In India, the government set a firm deadline - April 1, a month after the FT reported on the Israeli hacking of Tehran's camera system - by which Chinese cameras had to be banned in the country.

China is investing in new generations of cameras and software based on artificial intelligence, which can interpret scenes, recognize behavioral patterns and find footage based on written queries, to further sharpen its surveillance.

But the progress also opens up more opportunities for China’s rivals to effectively peer inside its borders. “They’re the ones putting up the cameras – all we have to do is find a way in. And there’s always a way in,” a security official from a member state of the Five Eyes intelligence alliance (the US, UK, Canada, Australia and New Zealand) told the FT.

However, such techniques are not foolproof. Opponents of states using AI for surveillance - particularly non-state actors such as militant groups Hezbollah and Hamas - often resort to low-tech methods, such as handwritten messages and old analog phone lines.

During the Gaza war, Israel's most sophisticated so-called "intercept, surveillance and reconnaissance" systems failed to locate Hamas leader Yahya Sinwar, the architect of the October 7, 2023 attack.

Israel and its allies poured enormous resources into the search for him, including “gait analysis” to try to identify the long stride of a tall, thin man in the crowd. He was eventually killed in October 2024, in what the military described as a chance encounter.