For most of its history, cybersecurity has been a contest between human attackers and human defenders, throttled on both sides by how fast people can work.That throttle is gone. Over the past two years, generative AI crossed from the defender's side of the table to the attacker's and the economics of an attack changed with it.The scale is easiest to grasp in one number. Cybercrime is projected to cost the world roughly $10.5 trillion in 2025, up from an estimated $6 trillion in 2021 and $3 trillion in 2015, according to Cybersecurity Ventures, a figure that, if it were a national economy, would rank third in the world. What is pushing that curve steeper is automation. In a survey of security professionals by SoSafe, 87% said their organization had encountered an AI-driven cyberattack in the past year, and 91% expected such threats to intensify over the next three. The most telling number was the one about defense: only about a quarter of teams were confident they could reliably detect these attacks.The texture behind those headlines is what makes this an inflection rather than a bad year. Phishing, still the most common intrusion vector, has been almost fully automated: independent analyses estimate that the large majority of phishing emails now use AI for drafting or obfuscation. Deepfake incidents have climbed more than 2,000% since 2022. And a growing share of detected malware is polymorphic, rewriting its own code on the fly to slip past signature-based defenses. Each of these maps to a threat class that barely existed as a mainstream enterprise concern three years ago: AI-powered phishing, adversarial machine learning, malicious use of generative AI, and automated, self-mutating malware.Defenders are responding with money. The market for AI-in-cybersecurity tools, the anomaly-detection engines, automated triage, and behavioral-analytics platforms that security teams now lean on, is forecast to roughly triple from $25–30 billion in 2024–25 to somewhere between $94 billion and $134 billion by 2030, depending on the analyst. But spending on tools has quietly run ahead of the ability to use them. The binding constraint in cybersecurity is no longer the software. It is the people who can wield it and here, in 2025, the story changed shape.The bottleneck moved from headcount to skillsFor the better part of a decade, the cybersecurity workforce conversation revolved around a single, dramatic figure: a global shortage of qualified professionals, most recently pegged by ISC2 at more than 4.7 million people. In its 2025 Cybersecurity Workforce Study, the field's largest, with 16,029 respondents, ISC2 did something quietly significant. It stopped publishing that gap number. Its reasoning: participants have consistently told the organization that the more pressing problem is no longer a shortage of people but a shortage of the right skills.The data behind that decision is stark. 95% of teams report at least one skills gap; 59% describe it as critical or significant up from 44% just a year earlier; and 88% say they suffered a security incident or operational failure tied to a skills shortage. Only 5% consider themselves fully resourced. And when ISC2 asked which technical skill was most needed, the answer, for the second consecutive year, was artificial intelligence — cited by 41% of respondents, ahead of cloud security. Two years ago, ISC2's own CISO has noted, AI was not even on the required-skills list.Buried in the same study is the mechanism that turns this from a hiring problem into an education problem. ISC2 flagged, in plain terms, that education and training are lagging behind the technology, that certification and course curricula have expanded but "may not be keeping pace with real-world demands." At the same time, hiring managers are chasing what the report calls "unicorns": individuals who are simultaneously strong in security fundamentals and capable of configuring, training, and interpreting AI systems. Those people barely exist. The sustainable path, the study concludes, is to build the skill into the people already in the field.That is the strategic backdrop against which a new kind of credential and a new kind of training pathway has appeared.AI security becomes its own disciplineThe clearest institutional signal that AI security has separated into its own discipline, rather than remaining a chapter inside a general security certification, arrived when CompTIA launched SecAI+ . It is CompTIA's first certification dedicated entirely to the intersection of AI and cybersecurity, and the first entry in a new "Expansion Series" explicitly designed to complement ,not replace, foundational credentials like Security+, CySA+, and PenTest+.What the exam chooses to weigh is itself an argument about what the discipline is. Its four domains are not balanced; they are deliberately tilted toward operational defense: Securing AI Systems at 40%, AI-Assisted Security at 24%, AI Governance/Risk/Compliance at 19%, and Basic AI Concepts at just 17%. In other words, the largest share of the exam is hands-on protection of machine-learning systems, adversarial-example defense, data-poisoning prevention, model-theft mitigation, pipeline hardening, anchored to recognized frameworks like the NIST AI Risk Management Framework, the OWASP Top 10 for LLMs, and MITRE ATLAS. This is a working skill set, not an awareness course. It is pitched at mid-career practitioners: CompTIA recommends three to four years in IT and two-plus years hands-on in security before attempting it.A certification is a market signal; a training pathway is how the workforce actually absorbs it. That is where the education layer comes in and where the practitioners building these programs offer a more grounded read than the threat statistics alone.The practitioner's view: risk, but also leverageBrad Smith, IT Certification Program Manager and instructor at the California Institute of Applied Technology (CIAT), is helping lead the rollout of one of the first US bootcamp pathways built around SecAI+. His framing is notable for what it doesn't do: it refuses the pure-threat narrative.While many organizations are focused on addressing the security risks associated with AI, fewer are taking a balanced approach that examines both how AI can introduce risk and how it can be leveraged to strengthen security operations.~ explains, Brad Smith.That duality is the heart of the discipline and, not coincidentally, the shape of the SecAI+ exam, which spends nearly a quarter of its weight on using AI in the security operations center, not just defending against it.Smith's most useful contribution, though, is where he points the risk. Asked which AI-driven threat is most pressing for the enterprises his graduates will join, he doesn't reach for the obvious answer of AI-powered phishing, sophisticated as that has become. He names something quieter: insider risk from AI adoption.This is the "shadow AI" problem, and it is the frontier most organizations underestimate: the enterprise data-protection challenge created not by an adversary but by well-meaning staff pasting confidential material into a chatbot. It squares with survey data showing that a majority of businesses have not fully implemented controls for their own in-house AI use, a governance gap sitting inside the productivity boom.On the demand side, Smith describes what employers are actually hiring against, and it is not a robot that replaces the analyst.Practitioners who can use AI to improve threat detection, triage large volumes of log data, establish behavioral baselines, and surface the anomalies that signal an incident. That matches the workforce data almost exactly: the same ISC2 study that ranked AI the #1 needed skill also found that 73% of professionals believe AI will create more specialized cybersecurity roles, not fewer.The pedagogy follows from that stance. CIAT built its program around the four SecAI+ domains, with hands-on labs, model fortification, adversarial-attack defense, data-poisoning prevention, AI-driven threat detection, meant to let students practice in realistic scenarios rather than memorize vocabulary. The ideal candidate, Smith says, mirrors CompTIA's own guidance: three to four years of cybersecurity experience and a foundational certification such as Security+ already in hand, "because the curriculum builds on established security knowledge." It is, in effect, the "targeted training and certification" pillar that workforce analysts at ISC2 and others keep naming as the way to close a skills gap that hiring alone cannot.The honest read: early, and worth watchingWhat is most credible about the practitioner view is where it declines to oversell. Pressed on the earnings premium a SecAI+ credential might command, Smith doesn't reach for a number. "Because SecAI+ is a newly launched certification," he says, "there is not yet sufficient market data to quantify a specific salary premium." That restraint is worth pausing on. Exam-prep marketers have already floated figures of a 15–30% premium for AI-security skills; those numbers are unverified and self-interested, and the more honest position, that the data isn't in yet, is the one a workforce educator should take. What can be said is directional, and it is what the rest of the evidence supports: a professional who pairs foundational security skills with demonstrated AI-security competence is positioned for a broader range of specialized roles than one who stays on the generalist track alone.There are real risks to watch. SecAI+ is months old; the labor-market data will take time to mature. And the very lag ISC2 identified, training curricula falling behind the technology, is a trap AI-security education could fall into just as easily, if programs teach the vocabulary of adversarial ML without the hands-on reps to actually defend a pipeline. The test for any pathway, CIAT's included, is whether graduates can do the work on day one, not merely name it.But the larger movement is not in doubt. The cybersecurity workforce story has inverted: the organizations that pull ahead won't be the ones that hire the most people, but the ones that upskill the people they already have into this dual competency fastest, defending AI systems, and fighting with them. A credential like SecAI+ and focused pathways like CIAT's are early instances of an entire field building a muscle it did not have eighteen months ago. Machine-speed attacks were always going to demand machine-age defenders. The surprise of the past year is how quickly "AI security" stopped being a specialty inside cybersecurity and started becoming a discipline of its own.Don't forget to like and share the story!