The evidence came from Flock Safety, a commercial automated video-surveillance system that had recorded Elzer\u2019s car. The officer refused to share details with the suspect, telling her to raise any objections in court.<\/p>\n
“You know we have cameras in that town. Not even a breath of fresh air will leave there without our knowledge,” the officer explained.<\/em><\/p><\/blockquote>\n
Convinced of her innocence, Elzer began gathering her own evidence. She had indeed been nearby that day \u2014 visiting a tailor \u2014 but had not stolen anyone\u2019s parcel.<\/p>\n
She compiled GPS data from apps on her phone and in her car, dashcam footage, witness statements and even photos of the clothes she wore on the day of the alleged crime.<\/p>\n
After repeated failed attempts to hand the information to the authorities, the suspect wrote directly to the police chief. He praised her work and said the summons had been annulled.<\/p>\n
As of December 2025, Flock Safety offered<\/a> access to 80,000 cameras across 49 American states.<\/p>\n
From monitor vans to crime prediction<\/h2>\n
Cameras on streets, in shops and in public buildings have long been ubiquitous. Today\u2019s smart cameras and data-processing methods, however, are a different beast.<\/p>\n
CCTV in the analogue era<\/h3>\n
Once upon a time, closed-circuit television (CCTV) meant a private network of cameras feeding a dozen monitors watched by a bored mall guard.<\/p>\n
The technology amounted to image sensors, screens and recording equipment.<\/p>\n
Security services have experimented with surveillance systems since at least the mid-20th century.<\/p>\n
Testing a CCTV system by British police in Trafalgar Square, 1960. Source: The National Archives<\/a>.\u00a0<\/figcaption><\/figure>\n In 1960 British police tested<\/a> two cameras monitoring Trafalgar Square during a visit by the king and queen of Thailand. Monitors sat in a van nearby. The trial exposed several technical issues and drew mixed reactions.<\/p>\n
In 1979 the government research outfit Police Scientific Development Branch developed<\/a> ANPR<\/span> using then-available optical character recognition methods.<\/p>\n
By the 1990s, cameras on intersections and fa\u00e7ades were the norm. Law enforcement folded CCTV and ANPR into everyday policing.<\/p>\n
Smart cameras<\/h3>\n
With miniaturised computing, pervasive connectivity and the rise of AI, the traditional CCTV rig has been supplanted by smart cameras tied to central databases and automated analytics.<\/p>\n
Such a device comes<\/a> with its own processor and OS, storage, and interfaces for local and internet connectivity, and sometimes a microphone for audio.<\/p>\n
A Flock Safety camera with ANPR functions. Source: Wikimedia<\/a>.\u00a0<\/figcaption><\/figure>\n Some makers embed<\/a> AI accelerators and NPU (Neural Processing Unit) modules to process data on the device in real time. Others offload AI tasks to external hardware.<\/p>\n
These systems can identify objects, recognise licence plates and faces, and store summaries of what they see. Capabilities vary with software configuration and vendor preferences.<\/p>\n
A brain behind the scenes<\/h3>\n
On its own, a smart camera can recognise objects and record their identifiers \u2014 a car\u2019s plate, a face, or a person\u2019s gait. An analytics hub aggregates camera feeds, fuses them with other sources and sends findings to an operator.<\/p>\n
Flock Safety offers<\/a> something similar in Nova, a \u201cpublic-safety data platform\u201d that ingests not only surveillance footage but also information from leaks, data-broker databases and other commercially available sources.<\/p>\n
Such a system builds dossiers with movement maps, preferences, browsing history, habits, police records and anything else available.<\/p>\n
Armed with such troves, AI can hypothesise about people\u2019s behaviour and alert operators to situations it deems suspicious. That option is already available<\/a> to Flock\u2019s customers.<\/p>\n
According to the company, Nova lets agencies close cases \u201cin one click\u201d.<\/p>\n
According to critics<\/a>, it allows warrantless tracking and invites sweeping privacy abuses.<\/p>\n
Colourful hair and code injection<\/h2>\n
Many people shrug at mass surveillance, seeing it as a tool for solving and preventing crime. Others are less sanguine about the bounds of personal freedom.<\/p>\n
The tug-of-war between smart cameras and those seeking privacy plays out on several fronts.<\/p>\n
Beyond public-policy battles, enthusiasts turn to camouflage \u2014 and to classic hacking.<\/p>\n
Spoofing<\/h3>\n
The most intriguing attacks on such devices are spoofing, or \u201cpresentation attacks\u201d, methods that manipulate the image a camera receives.<\/p>\n
These include masks, retroreflectors, specialised textures and other ways to \u201cspoil\u201d an image so the system fails to detect or correctly identify an object.<\/p>\n
In 2016 designer Scott Urban\u2019s Reflectacles<\/a> project offered eyewear with a retroreflector that bounces a surveillance camera\u2019s infrared light back, blowing out the face image.<\/p>\n
Reflectacles on surveillance video. Source: Kickstarter.<\/figcaption><\/figure>\n This brute-force tactic can deprive a single camera of usable data, but it falters when multiple angles are in play.<\/p>\n
Berlin-based researcher and artist Adam Harvey developed a series of CV Dazzle<\/a> countermeasures to facial recognition.<\/p>\n
His early-2010s looks relied on asymmetric hairstyles and makeup elements designed to fool the then-popular Viola\u2013Jones algorithm<\/a>, which detects under-eye and nasal shadows, facial symmetry and nose-bridge location.<\/p>\n
His answer: unconventional shadow placements and colours that contrast with skin tone.<\/p>\n
CV Dazzle Look 5. Source: Adam.harvey.studio.<\/figcaption><\/figure>\n With the rise of AI facial-recognition systems, the old tricks dated; in 2020 Harvey proposed an updated makeup approach.<\/p>\n
CV Dazzle Look 6 and 7. Source: Adam.harvey.studio.<\/figcaption><\/figure>\n He stressed he was demonstrating a technique, not fixed patterns; the best solution depends on the surveillance context.<\/p>\n
Similar methods apply to plate recognition. American enthusiast Benn Jordan described<\/a> how to craft \u201cadversarial\u201d textures for ANPR detectors.<\/p>\n
Using open recognition models, Jordan trained a neural network to generate visual noise that, when overlaid on a licence plate, makes the model read incorrect characters or fail to \u201csee\u201d the plate at all.<\/p>\n
The problem with visual tricks is their fragility. Efficacy depends on conditions and on camera coverage. Meanwhile, surveillance vendors keep expanding the cues used for recognition, such as a person\u2019s distinctive gait or a car\u2019s colour and unique modifications.<\/p>\n
Researchers continue to probe advanced models for weaknesses, but a clearer threat to smart-camera systems comes from hackers.<\/p>\n
Device compromise and network attacks<\/h3>\n
Like any internet-connected computer \u2014 AI or not \u2014 smart cameras and their back-end infrastructure are vulnerable to hacking.<\/p>\n
Over the years, numerous vulnerabilities have been documented.<\/p>\n
In 2021 a code-injection vulnerability<\/a> was found in Hikvision surveillance cameras. It allowed full device takeover, software installation and access to other cameras on the network.<\/p>\n
In 2023 the operating system of Axis smart cameras contained a flaw<\/a> enabling arbitrary command execution during installation of ACAP applications<\/a>.<\/p>\n
In 2025 surveillance systems by Dahua were found to have two vulnerabilities<\/a> related to remote command execution and buffer overflow. Both allowed an attacker full control of a camera.<\/p>\n
A separate attack vector is direct interaction with hardware often mounted outdoors in public places. An attacker can use service interfaces, access local storage or modify a device for their own ends.<\/p>\n
To counter direct attacks, manufacturers encrypt data, use hardware verification of firmware<\/a> and add cryptographic signatures<\/a> to video files.<\/p>\n
A properly configured device cannot simply be \u201creflashed\u201d or have data pulled off it. But mistakes happen.<\/p>\n
In 2025, 404 Media reported<\/a> that at least 60 Flock Safety Condor AI cameras with people-tracking features were left open to unauthorised access.<\/p>\n
Cybersecurity specialist John Gaines and the above-mentioned researcher Benn Jordan found the devices\u2019 IP addresses using the Shodan search engine and discovered they could connect with no login or password.<\/p>\n
A journalist in the frame of a Flock surveillance camera. Source: 404 Media.<\/p>\n
Anyone could watch live streams, download 30 days of archives, change settings and read system logs.<\/p>\n
The vendor blamed \u201ca misconfiguration affecting a limited number of devices\u201d and said the problems had been fixed.<\/p>\n
The same researchers said another Flock camera model exposes an open Wi-Fi access point if certain buttons on the case are pressed, granting full control over the device and its software.<\/p>\n
Gaines published an analysis of these and other flaws in Flock\u2019s system in a separate document<\/a> covering 55 items.<\/p>\n
In its official response the company said<\/a> the issues were already known, and that would-be attackers rely on direct access to cameras and \u201cdeep knowledge of the equipment\u2019s internals\u201d.<\/p>\n
The vendor stressed that all necessary updates are delivered without customer action and that system operations face no threat.<\/p>\n
Tackling the \u201cpartially competent\u201d<\/h2>\n
Automated surveillance systems, especially with AI, have become a convenient tool for law enforcement.<\/p>\n
Vendors sell their wares on promise: here\u2019s the suspect\u2019s car and a map of its movements; here\u2019s the address. Cases can now be closed with a click.<\/p>\n
That ease is habit-forming. People tend to over-rely on automated outputs \u2014 a classic cognitive bias<\/a>.<\/p>\n
AI-powered automation follows the same pattern: many users assume ChatGPT\u2019s answers are correct and ignore contradictions. In everyday contexts this can distort perceptions and, in some cases, lead to psychoses<\/a>.<\/p>\n
Even under an engineering ideal and in the hands of authorised operators, a large AI-driven surveillance system can do harm.<\/p>\n
![\"image\"](\"https:\/\/forklog.com\/wp-content\/uploads\/img-c767755fecab2bbf-7851533659120250.webp\")
![\"image\"](\"https:\/\/forklog.com\/wp-content\/uploads\/img-2b5d664375b0a2ec-7851667301706727.webp\")
![\"image\"](\"https:\/\/forklog.com\/wp-content\/uploads\/img-20bd2f2e90c1aa4b-7851533650200844.webp\")
![\"image\"](\"https:\/\/forklog.com\/wp-content\/uploads\/img-782e2aba257ab4c9-7851533641910280.webp\")
![\"image\"](\"https:\/\/forklog.com\/wp-content\/uploads\/img-f2781be8b803167c-7851533845101894-1024x770.png\")