Tag: penetration testing

  • Cyber resilience act compliance is already broken by AI agents

    Cyber resilience act compliance is already broken by AI agents

    Table of contents

    Cyber Resilience Act compliance feels like solid ground once you’ve passed the audit and picked up your CE mark. A new study argues that ground already gave way, because AI agents can turn a fully certified device into an open door without touching a single line of its code.

    The Hookii robotic lawn mower passed every check the EU regulation asks for, and so did the Unitree G1 humanoid robot. Both are certified and legal to sell across the bloc, at least on paper. In March 2026, an AI agent costing about as much as a coffee run took over a fleet of 267 of those mowers without breaking a single law or touching the factory floor. The certificate is still hanging on the wall, but reality already outran it.

    What Cyber Resilience Act compliance actually demands

    The EU Cyber Resilience Act reaches full force in December 2027 with a compliance loop at its center. Manufacturers assess risk and handle whatever flaws surface, then ship patches on adeclared schedule while reporting active exprs. The premise underneath is that softwarealways ships with flaws and only needs managing.

    That design rests on four unstated assumptions about the real world, and each one has to hold for the process to meaanything.

    1. Finding a flaw takes real expertise and real time, so only a small, slow-moving set of vulnerabilities becomes known at any given moment.
    2. A product’s full set of exploitable bugs can be mapped out by the day it ships, and that map stays roughly accuraafterward.
    3. Exploitation is rare enough, and visible enough, that spotting an incident counts as a meaningful signal.
    4. Remediation moves faster than attackers do, so a scheduled patch cycle is enough to stay ahead.

    Víctor Mayoral-Vilches of Alias Robotics, the paper’s author, dates the mismatch to a ten-week window. The European Commission drafted the bill in September 2022 and ChatGPT shipped that November, freezing the regulation’s picture othe world at the exact moment that picture s
    ## Why the volume argument misses the point

    Most commentary on AI and cybersecurity fixates on a single number, that agents find more bugs than people do. True,but that’s the smaller half of the story forence Act compliance.
    When a GPT-4 agent exploited 87% of freshly es in April 2024 given the CVE description,that alone is survivable. Companies re-priormost and documenting the risk they accept onthe rest, which is exactly the behavior Article 14 anticipated when it limited mandatory reports to bugs under active attack and left routine scanner findings outside the count. This is no lab curiosity either, since AI hacking tools are already probing production servers around the clock.

    The clock does far more damage than the raw count, though. Median time from disclosure to weaponization stood at 771 days back in 2018; by 2023 it was 5.3 days, an exponential decline fitting at R²=0.98 that sat near zero in 2025, while the share of bugs weaponized before or at public disclosure climbed from 19% up to 54%.

    A certificate that lies without the product changing

    Cyber resilience act compliance

    Cyber Resilience Act compliance turns fragilrequirement that products ship “without known exploitable vulnerabilities.” That clause assumes “known” on day one roughly equals “findable” a week later.


    Google’s Big Sleep agent rediscovered a hidden SQLite flaw on demand in November 2024, and running that kind of attack today costs roughly a hundred dollars a try, yet the product and its certificate stayed exactly the same on paper. Its real exploitable footprint moved anyway, because the conditions around

    That hits every manufacturer shipping a networked device into the EU, whether it’s an IP camera, an industrial
    controller, a warehouse robot, or a smart thzen and fully compliant can become an opendoor a week after certification, because the flaw grew out of the attacker’s toolkit, sitting entirely outside the product’s own code.

    Two robots, one proof

    The paper backs its claim with two devices tested directly under Cyber Resilience Act scope, the Unitree G1 humanoid
    and the Hookii mower.

    Undefended, an AI agent rooted the G1 through a Bluetooth command-injection bug, exploiting an identical AES key baked into every unit in the fleet. From there it decrypted the robot’s telemetry and reached teleoperation, with a success rate of 79%. On the mower, 38 chained vulnerabilities bypassed the safety geofence across a fleet of 267 devices at 75% success.

    Enroll both robots in the Robot Immune System, an autonomous defensive AI agent, and attacker success collapses to 14% and 8%, with detection and containment landing under 8 and 12 seconds. The paper’s conclusion cuts both ways, because attacker and defender run on the same technology, so the only certification that holds up is one that never stops running. I covered the mechanics behind that idea in how autonomous cyber defense learns.

    Ways to pressure-test your compliance program

    If you advise clients on Cyber Resilience Act compliance or ship connected products into the EU, these moves matter more than the paperwork.

    1. Schedule a re-test after certification lands, since a point-in-time audit only tells you about the day it ran.
    2. Budget for continuous monitoring as a recurring cost, because the disclosure-to-exploit window is now measured in days, sometimes hours.
    3. Treat a defensive AI agent as core infras Act compliance; the paper’s own data showsattacker success collapsing from 79% to 14% once one is running.
    4. Ask any vendor how they detect drift between what was certified and what’s actually running today.

    December 2027 is when the Cyber Resilience Act switches on in full, certifying products against a world that has already moved on. A once-a-year compliance stamp buys nothing past the following Tuesday, and continuous, agent-operated defense is the only way left for longer than a week.

    Source: https://arxiv.org/abs/2607.07109


    Want More? Subscribe to The Dossier

    Every week in your inbox:

    📡 THE INTELLIGENCE FEED – 3-5 curated links: [Research] [Policy] [Tools] [Incidents]
    💡 ONE ADVICE – One actionable AI/cybersecurity tip you can use today

  • AI Hacking Tools Are Trying to Breach Your Servers. Here is Why They Fail.

    Table of contents

    Picture this. You hand the keys to your entire server infrastructure to a supercomputer, hoping modern AI hacking tools will find every flaw. You point at a known vulnerability in your code and give it one command. Exploit this bug.

    Prove to me this system can be compromised. You wait for the fireworks. Instead, the multi-billion-dollar brain stutters. It spits out gibberish errors. Then it completely gives up. Researchers from UC Berkeley just exposed this exact scenario.

    The Illusion of the All-Knowing Machine

    We have been spoon-fed a narrative about omnipotent algorithms tearing through Pentagon firewalls in seconds. Media outlets pumped up the hysteria. Tech companies started firing junior security analysts. They actually believed basic scripts could handle the heavy lifting.

    The truth turned out to be far more brutal and embarrassing for the developers of AI hacking tools. The creators of CyberGym built a massive testing ground. They gathered over 1,500 real-world vulnerabilities across nearly 200 software projects. The task was deceptively simple. The machine had to generate a working Proof-of-Concept exploit based on a text description and the codebase.

    ai hacking tools

    The top-performing models on the market hit a massive brick wall. They achieved a success rate of roughly 20 percent. Eight out of ten attempts ended in total failure. This completely destroys the hype about machines stealing jobs from seasoned penetration testers. These systems can spit out thousands of lines of syntactically perfect code. Deep comprehension, however, completely eludes them.

    Why the Industry Has It Completely Backwards

    Most people view technological progress through the lens of static benchmarks. A machine passes a medical exam. We automatically assume it can handle a dynamic network environment. This is a massive cognitive bias. An exam operates within a closed, predictable set of rules. Hacking is the exact opposite. Hacking requires you to break the rules. We constantly confuse raw processing speed with actual cunning. When evaluating AI hacking tools, we must look at actual performance, not theoretical capacity.

    Most AI hacking tools look at code and predict the next token based on statistical probabilities. They do not actually understand the logic. They fail to grasp that altering a single variable in an obscure module will cause a cascading memory failure on a separate server. Second-order consequences remain completely out of reach for current architectures.

    You cannot feed a machine millions of server logs and expect it to magically develop a predator’s instinct. Imagine a burglar. He knows what a lock looks like. He can describe its internal mechanism in a hundred languages. When you hand him a lock pick, he tries to shove the instruction manual into the keyhole.

    The Real-World Fallout for Your Business

    What does this mean for a company founder or an IT director? It creates a dangerous false sense of security. If you rely strictly on AI hacking tools to audit your codebase, you leave your company-wide open to attack. These bots will catch typos. They will flag basic misconfigurations, like an exposed AWS bucket. They will fail completely against complex zero-day vulnerabilities.

    Let’s look at a concrete scenario. You launch a new payment processing app. You hire an automated bot to scan the 50,000 lines of code. The bot gives you a clean bill of health. You push the app to production. A week later, criminals drain $250,000 from user accounts. Why? A human hacker found a race condition.

    They sent two withdrawal requests in the exact same millisecond. The algorithm never even considered simulating server load physics against CPU timing constraints. The bot just read clean text. The human read between the lines.

    CyberGym researchers did discover 35 new vulnerabilities. They also found 17 incomplete patches. The machine did not do this alone. It simply fetched the right diagnostic data for human operators. You fire the humans, the software just gathers dust.

    Companies putting blind faith in AI hacking tools will become the easiest targets on the internet. Criminals know exactly where the algorithms have blind spots. They will strike exactly there.

    Under the Hood of an AI Breakdown

    Let’s cut to the chase and look at the technical mechanics. Why do these systems fail at writing exploits? The core issue is context management. Executing a successful attack requires maintaining a complex state over multiple steps. You must send a malformed data packet. You must wait for a highly specific response. You must hijack the instruction pointer in active memory.

    Models get lost in long chains of cause and effect. They hallucinate non-existent functions. They try to use methods patched out in 2018. Not only that, but they lack the ability to correct course dynamically. A human sees a segmentation fault and immediately analyzes the memory dump. A machine sees the exact same error and enters an infinite loop.

    It tries the exact same broken command over and over again. Writing a buffer overflow exploit requires precise calculations of memory offsets. Current models just guess. They throw random strings at the wall. They blindly hope something breaks. Direct interaction with a live execution environment exposes every single weakness of AI hacking tools.

    Securing Your Systems the Right Way

    Stop treating these programs like magic. Keep your senior engineers on payroll. Automated scanning applications and AI hacking tools are nothing more than noisy toys in the hands of amateurs right now. Handing the keys of your kingdom to an algorithm is an open invitation for disaster.

    If you want to protect your network today, implement these three mandatory protocols:

    1. Schedule manual penetration tests every 6 months using certified human security engineers.
    2. Deploy AI hacking tools strictly for initial static code analysis, but never rely on them for final production sign-off.
    3. Install multi-layered monitoring systems that detect behavioral anomalies rather than relying on known exploit signatures.

    The real war for your data will continue to be fought by human minds.


    Want More? Subscribe to The Dossier

    Every week in your inbox:

    📡 THE INTELLIGENCE FEED – 3-5 curated links: [Research] [Policy] [Tools] [Incidents]
    💡 ONE ADVICE – One actionable AI/cybersecurity tip you can use today