SMS Blog

Medieval Sword or Modern Drone: How to Battle “Infinity Script Kiddies”

Introduction

There’s a battle already underway, and the enemy isn’t a lone mastermind, it’s a swarm. AI has unleashed “infinity script kiddies”, endless, automated adversaries that probe every weakness at machine speed, all at once, all the time. Defenders are getting faster too, but speed alone won’t win this fight. To survive it, you’ll need two weapons at once: a medieval sword and a modern drone. The sword is your foundational security, the timeless, unglamorous fundamentals that have always kept defenders alive. The drone is your modern, AI-enabled defense, fast and far-seeing, built to meet attackers at machine speed. Neither one wins alone.

To see why we need both, we first have to be honest about how the battlefield has actually changed. Three uncomfortable truths define the new terrain. First, you can’t patch your way to secure software; second, attackers don’t break in, they log in; and last, AI has turned yesterday’s amateurs into tireless, relentless threats. Each one reshapes where our defenses hold and where they crumble, and each one shows why the sword and the drone need each other.

But naming the threat isn’t the same as defeating it. From there, we point toward a way through. It’s not a single silver bullet, but a direction. We pair the drone’s modern, AI-enabled defenses with the sword’s timeless fundamentals that keep us standing long enough to use them.


You can’t patch your way to secure software

Let’s be clear: patching is essential, and it isn’t going away. The trouble is treating it as the entire strategy. AI is dramatically accelerating the discovery of vulnerabilities, compressing what once took years into a far shorter window, and it’s doing that on both sides of the battle. In 2024, Google’s Big Sleep, an AI agent, discovered an exploitable memory flaw in SQLite before it reached production; the issue had been missed for years by traditional methods. This discovery is a good illustration of how AI can easily find those long-hidden flaws, considerably narrowing the window between hidden bug and exploitation. Britain’s National Cyber Security Centre recently warned organizations to prepare now for a coming “patch wave” , essentially a rush of urgent software updates across the whole technology stack, triggered as AI tools in skilled hands surface large numbers of hidden flaws in quick succession. And the pool of latent bugs is enormous. Decades of accumulated “technical debt” sit in our infrastructure, and AI is poised to drain that reservoir fast.

That is exactly why patching alone can’t save you. When fixes become a continuous firehose, “patch faster” comes with a hidden bill of more frequent change, more brittle dependencies, and more opportunities for supply-chain compromise. The patch itself can become a new attack surface, and the operational burden of emergency updates turns into its own systemic risk.

The answer isn’t to patch less; instead, it’s to depend on patching less. That means building security in from the start rather than bolting it on after disclosure. That looks like secure software development practices, rigorous testing, and risk-based vulnerability prioritization so teams fix what actually matters first instead of drowning in every new CVE. It also means supply-chain governance: knowing what components are in your software, vetting dependencies, and demanding the same rigor from your vendors. This is where the drone comes in. That same AI that finds flaws for attackers can find them for you first. AI-assisted code review, automated penetration testing, and AI-driven vulnerability prioritisation let teams find and rank those weaknesses before the patch wave hits. Patch quickly when the wave hits, yes, but build so that a single missed patch isn’t the line between secure and breached.


Hackers don’t break in, they log in

Modern attackers increasingly don’t need exotic exploits; they need valid access. If credentials, identity proofing, and verification workflows are weak, the attacker’s job becomes simple. The attacker only needs to show up as “the right person,” get handed the keys, and walk through the front door. In thatworld, “preventing break-ins” matters less than preventing impersonation, stopping unauthorized account use, and detecting behavior that doesn’t match the real user, because the system will happily authenticate an impostor.

Consider business email compromise (BEC), one of the costliest attacks in the world. There’s rarely a dramatic breach. Instead, an attacker takes over or convincingly spoofs a trusted email account and simply asks for a wire transfer, for a change of banking details, or for an urgent payment. The request “logs in” through legitimate trust, and the money walks out the front door.

The FBI recently documented the same pattern in the physical world, as noted in The Record. Cybercriminals stole roughly \$725 million in cargo across the U.S. and Canada in a single year, not by hijacking trucks but by impersonating legitimate brokers and carriers. They phished their way into load-board systems, posted fraudulent listings using compromised carrier accounts, and even changed a legitimate carrier’s official contact and insurance details with federal regulators. They then had unwitting drivers reroute shipments straight to them. No forced entry required; they logged in and someone handed over the goods.

The defensive lesson follows directly. Assume passwords will be phished and focus on controls that survive it. Phishing-resistant MFA, hardware security keys and passkeys built on FIDO2/WebAuthn which bind authentication to the real site and can’t be relayed by a fake login page, closes the door that ordinary passwords and even one-time codes leave open. Pair it with strong identity proofing, least-privilege access, out-of-band verification for high-risk requests like payment or banking changes, and behavioral monitoring that flags a “valid” user doing invalid things. The power of using AI as a tool with the more traditional weapons comes in the different information it can react to. An AI-powered behavioral analytics tool can build a baseline per individual. AI can then interrupt lateral movement mid-session, even though the account was authenticated cleanly, when the account simply begins accessing servers they have never touched before. None of these tools can be dispensed with; they are all needed for this battle.


AI models are “infinity script kiddies”

First, the term. A “script kiddie” is the oldest, least-respected type of attacker. They’re an inexperienced hacker who copies and pastes code they don’t actually understand and run pre-built tools in a “spray and pray” fashion, hoping to get lucky. Historically they weren’t much of a threat. Their off-the-shelf tools had well-known signatures that standard defenses caught easily.

Now add AI, and multiply. “Infinity script kiddies” describes what happens when generative AI hands those same low-skill attackers near-unlimited capability and near-unlimited attempts. AI writes the code they couldn’t, tunes attacks they didn’t understand, and lets them iterate at machine speed, probing, phishing, guessing, and refining social engineering endlessly until something works. The “infinity” is the point. It’s not one gifted adversary, but an unbounded swarm of mediocre ones, each cheap to spin up and relentless in volume.

This isn’t hypothetical anymore. Evidence is building that attackers are using AI to enable their attacks. On a criminal forum, a threat actor known as KuroCracks claimed to have used ChatGPT to optimise a scanner tool they developed targeting CVE-2024-10914. This was an unauthenticated remote codeexecution vulnerability in D-Link NAS devices, per S2W. These devices are end of life and will never be patched, yet tens of thousands remain exposed online. This is exactly the type of forgotten-asset environment where script kiddies thrive. Kela’s dark web monitoring found a 219% increase in malicious AI mentions. And because AI takes attackers who should be harmless and lets them punch far above their weight, this is going to rapidly become a problem.

The danger was never the skill of any single kiddie; it’s that AI gives an infinite number of them the reach of an expert, pressuring every weak control simultaneously, all the time. AI defenders built on behavioral pattern recognition, vs known signatures, are already being deployed to meet these challenges. They can catch AI-generated phishing lures, along with novel attack variants, because they evaluate context and intent, instead of simply looking at fingerprints.


Main point: The question we’re avoiding

So how can anyone defend themselves against all this, against a world where patching is a risk, “logging in” is the primary breach path, and attackers have effectively unlimited retries?

Before the answer, a harder question: why do so many organizations keep losing this fight while spending more than ever? Because we’re wired to buy our way out. New technology is tangible, board-friendly, and exciting. A shiny “next-gen AI platform” is far easier to fund than the unglamorous work of rotating credentials, enforcing MFA everywhere, and decommissioning a forgotten server. Vendors sell outcomes in a logo; fundamentals are sold in sweat. So budgets tilt toward the new tool, and a quiet assumption creeps in that the last purchase must have already covered the basics.

It rarely did. Foundational security gets neglected precisely because it’s boring, continuous, and never “done.” Asset inventory, patch discipline, least privilege, backups, and awareness training earn no headlines when they work and deliver no dopamine when you fund them. They compete against the emergency of the week and lose. Yet year after year, breach investigations land on the same unglamorous causes. There were stolen or weak credentials, unpatched known bugs, misconfigurations, and a person who got tricked, not exotic zero-days. That’s why practitioners keep preaching a return to the basics, and why the security pendulum is swinging back toward preventative controls. The attackers keep finding the open doors first.

Here’s the trap AI sets. Because “infinity script kiddies” move at machine speed, it’s tempting to answer a machine with a machine. One can buy an AI defender and call it done. But AI is an amplifier, not a substitute. Layer world-class anomaly detection on top of shared admin passwords and a flat network, and you’ve built a very expensive smoke alarm over an unlocked house. The AI will detect the intrusion faster while the attacker still strolls in. Point that same AI at a hardened environment, MFA enforced, least privilege in place, assets known and patched, and it becomes a genuine force multiplier, catching the subtle, mutating attacks that signatures miss. Hugging Face’s July 2026 breach shows this in practice. An autonomous AI agent got in through a malicious dataset, harvested credentials, and moved laterally across internal systems over a weekend. The Hugging Face team conducting the breach analysis were able to use AI to access more than 17,000 recorded attacker events, compressing days of forensic reconstruction into hours. AI raises the ceiling of your defense;

fundamentals set the floor.

So the answer is two-fold. You need both new-fangled weapons and old-fashioned fundamentals. You need the “drone”: modern, AI-enabled defenses like behavioral anomaly detection, automated hygiene, and AI-assisted testing that can operate at machine speed. And you need the “sword”: foundational controls like MFA, strong authentication practices, and workforce awareness that stop credential-based intrusion and social engineering before “infinity script kiddies” get traction. Either one alone fails. The drone without the sword is fast but hollow; the sword without the drone is sturdy but outpaced.


Conclusion

Cybersecurity in the AI age is not a choice between tradition and innovation. It’s a requirement to master both. The drone helps you see, respond, and adapt at scale. The sword keeps your core posture intact when attackers try the simplest path of identity, access, and human trust. If you want to win modern-day cybersecurity battles, build for the future, but reinforce the fundamentals that keep you alive long enough to get there.

So here’s the challenge. Before you fund the next shiny tool, answer three unglamorous questions:

  • Can you name every asset you own?
  • Is Phishing Resistant MFA truly enforced everywhere?
  • Would your people catch a convincing phish today?

Those “back to basics” answers, not the logo on your latest platform, are what will decide the outcome. The organizations most prepared for AI won’t necessarily have only the latest tools. They’ll be the ones that combine modern automation with disciplined execution of the security fundamentals that have always mattered. Master both the sword and the drone, and you won’t just survive this fight against the infinity script kiddies. You’ll be ready for the next one.


Bibliography

  1. Risky Business. “Risky Business #832 — Anthropic unveils magical 0day computer God.” https://risky.biz/RB832/
  2. Risky Business. “Risky Business #836 — You can’t patch the bugpocalypse.” https://risky.biz/RB836/
  3. Reuters. “US officials weigh cutting deadlines to fix digital flaws amid worries over AI.” https://www.reuters.com/legal/litigation/us-officials-weigh-cutting-deadlines-fix-digital-flaws-amid-worries-over-ai-2026-05-01/
  4. The Record. “British cyber agency warns of looming ‘patch wave’ as AI speeds flaw discovery.” https://therecord.media/british-cyber-ai-patch-wave
  5. The Record. “Hackers earning millions from hijacked cargo, FBI says.” https://therecord.media/hackers-earning-millions-from-hijacked-cargo-fbi
  6. Medium. “Are We Ready for the Script Kiddies of AI?” https://medium.com/data-science/are-we-ready-for-the-script-kiddies-of-ai-754b23f5fcf8
  7. Exabeam. “The Rise of Script Kiddies in an AI World.” https://www.exabeam.com/blog/ten18/the-rise-of-script-kiddies-in-an-ai-world/
  8. NISC. “Back to the Basics: Simple Cybersecurity Practices That Still Very Much Matter.” https://www.nisc.coop/blog/back-to-the-basics-simple-cybersecurity-practices-that-still-very-much-
  9. matter/
  10. Kela. “2025 AI Threat Report.” https://info.ke-la.com/hubfs/Reports/KELA Report – 2025 AI Threat Report.pdf
  11. The Hacker News. “Google’s AI Tool Big Sleep Finds Zero-Day Vulnerability in SQLite Database Engine.” https://thehackernews.com/2024/11/googles-ai-tool-big-sleep-finds-zero.html
  12. ISACA. “The Need For AI Powered Cybersecurity to Tackle AI Driven Cyberattacks.” https://www.isaca.org/resources/news-and-trends/isaca-now-blog/2024/the-need-for-ai-powered-cybersecurity-to-tackle-ai-driven-cyberattacks
  13. Hugging Face. “Security incident disclosure — July 2026.” https://huggingface.co/blog/security-incident-july-2026
  14. CSO Online. “Hugging Face breach shows why incident response needs multi-model AI.” https://www.csoonline.com/article/4201361/hugging-face-breach-shows-why-incident-response-needs-a-multi-model-ai-strategy.html
Picture of Marija Strazdas

Marija Strazdas

Marija is captivated by logic and technology, and blends that with a quest for understanding how we humans use technology, and what that means for our human society. She has worked as a Solution Architect in Managed Hosting and Cybersecurity, across businesses ranging from AWS to a startup. View Marjia's LinkedIn

Leave a Reply

Your email address will not be published. Required fields are marked *