In early 2024, a putting deepfake fraud case in Hong Kong introduced the vulnerabilities of AI-driven deception into sharp reduction. A finance worker was duped throughout a video name by what seemed to be the CFO—however was, in actual fact, a complicated AI-generated deepfake. Satisfied of the decision’s authenticity, the worker made 15 transfers totaling over $25 million to fraudulent financial institution accounts earlier than realizing it was a rip-off.
This incident exemplifies extra than simply technological trickery—it alerts how belief in what we see and listen to might be weaponized, particularly as AI turns into extra deeply built-in into enterprise instruments and workflows. From embedded LLMs in enterprise techniques to autonomous brokers diagnosing and even repairing points in stay environments, AI is transitioning from novelty to necessity. But because it evolves, so too do the gaps in our conventional safety frameworks—designed for static, human-written code—revealing simply how unprepared we’re for techniques that generate, adapt, and behave in unpredictable methods.
Past the CVE Mindset
Conventional safe coding practices revolve round identified vulnerabilities and patch cycles. AI modifications the equation. A line of code might be generated on the fly by a mannequin, formed by manipulated prompts or information—creating new, unpredictable classes of danger like immediate injection or emergent conduct outdoors conventional taxonomies.
A 2025 Veracode research discovered that 45% of all AI-generated code contained vulnerabilities, with widespread flaws like weak defenses in opposition to XSS and log injection. (Some languages carried out extra poorly than others. Over 70% of AI-generated Java code had a safety problem, for example.) One other 2025 research confirmed that repeated refinement could make issues worse: After simply 5 iterations, crucial vulnerabilities rose by 37.6%.
To maintain tempo, frameworks just like the OWASP High 10 for LLMs have emerged, cataloging AI-specific dangers comparable to information leakage, mannequin denial of service, and immediate injection. They spotlight how present safety taxonomies fall brief—and why we’d like new approaches that mannequin AI risk surfaces, share incidents, and iteratively refine danger frameworks to mirror how code is created and influenced by AI.
Simpler for Adversaries
Maybe probably the most alarming shift is how AI lowers the barrier to malicious exercise. What as soon as required deep technical experience can now be completed by anybody with a intelligent immediate: producing scripts, launching phishing campaigns, or manipulating fashions. AI doesn’t simply broaden the assault floor; it makes it simpler and cheaper for attackers to succeed with out ever writing code.
In 2025, researchers unveiled PromptLocker, the primary AI-powered ransomware. Although solely a proof of idea, it confirmed how theft and encryption may very well be automated with an area LLM at remarkably low price: about $0.70 per full assault utilizing industrial APIs—and primarily free with open supply fashions. That sort of affordability may make ransomware cheaper, quicker, and extra scalable than ever.
This democratization of offense means defenders should put together for assaults which are extra frequent, extra diversified, and extra artistic. The Adversarial ML Menace Matrix, based by Ram Shankar Siva Kumar throughout his time at Microsoft, helps by enumerating threats to machine studying and providing a structured method to anticipate these evolving dangers. (He’ll be discussing the issue of securing AI techniques from adversaries at O’Reilly’s upcoming Safety Superstream.)
Silos and Talent Gaps
Builders, information scientists, and safety groups nonetheless work in silos, every with completely different incentives. Enterprise leaders push for speedy AI adoption to remain aggressive, whereas safety leaders warn that shifting too quick dangers catastrophic flaws within the code itself.
These tensions are amplified by a widening expertise hole: Most builders lack coaching in AI safety, and lots of safety professionals don’t totally perceive how LLMs work. Consequently, the previous patchwork fixes really feel more and more insufficient when the fashions are writing and operating code on their very own.
The rise of “vibe coding”—counting on LLM solutions with out evaluation—captures this shift. It accelerates growth however introduces hidden vulnerabilities, leaving each builders and defenders struggling to handle novel dangers.
From Avoidance to Resilience
AI adoption gained’t cease. The problem is shifting from avoidance to resilience. Frameworks like Databricks’ AI Danger Framework (DASF) and the NIST AI Danger Administration Framework present sensible steering on embedding governance and safety immediately into AI pipelines, serving to organizations transfer past advert hoc defenses towards systematic resilience. The objective isn’t to eradicate danger however to allow innovation whereas sustaining belief within the code AI helps produce.
Transparency and Accountability
Analysis reveals AI-generated code is usually less complicated and extra repetitive, but in addition extra weak, with dangers like hardcoded credentials and path traversal exploits. With out observability instruments comparable to immediate logs, provenance monitoring, and audit trails, builders can’t guarantee reliability or accountability. In different phrases, AI-generated code is extra prone to introduce high-risk safety vulnerabilities.
AI’s opacity compounds the issue: A perform could seem to “work” but conceal vulnerabilities which are tough to hint or clarify. With out explainability and safeguards, autonomy shortly turns into a recipe for insecure techniques. Instruments like MITRE ATLAS may help by mapping adversarial techniques in opposition to AI fashions, providing defenders a structured method to anticipate and counter threats.
Trying Forward
Securing code within the age of AI requires greater than patching—it means breaking silos, closing ability gaps, and embedding resilience into each stage of growth. The dangers could really feel acquainted, however AI scales them dramatically. Frameworks like Databricks’ AI Danger Framework (DASF) and the NIST AI Danger Administration Framework present buildings for governance and transparency, whereas MITRE ATLAS maps adversarial techniques and real-world assault case research, giving defenders a structured method to anticipate and mitigate threats to AI techniques.
The alternatives we make now will decide whether or not AI turns into a trusted companion—or a shortcut that leaves us uncovered.
Guarantee your techniques stay safe in an more and more AI-driven world
Be part of Chloé Messdaghi and a lineup of prime safety professionals and technologists for O’Reilly’s Safety Superstream: Safe Code within the Age of AI. They’ll share sensible insights, real-world experiences, and rising developments that may make it easier to code extra securely, construct and deploy safe fashions, and defend in opposition to AI-specific threats. It’s free for O’Reilly members. Save your seat right here.
Not a member? Join a free 10-day trial to attend—and take a look at all the opposite nice sources on O’Reilly.
