Skip to content
Securafy AI Lab by Securafy
AI Security & Governance

AI Security Enrollments Are Up 358%. Is Your Team's Training Keeping Up?

Coursera data shows US AI security enrollments up 358% year over year, while ISC2 and ISACA research shows security leaders prioritizing AI and many organizations unsure how to halt a misbehaving AI system. Rodney Hall explains why AI security is now an end-user skill and how to pair training with the platform controls it depends on.

In this article

If one of your employees connected an AI assistant to the shared drive tomorrow and it started doing something nobody asked for, would anyone on your team know how to stop it? That is the question underneath one of the sharpest training shifts of 2026. Learners are moving into AI security courses far faster than into AI courses in general, and security leaders now rank AI as their top training need. Most small and mid-sized teams have not adjusted their own AI security training to match.

What does the 358% jump in AI security enrollments actually measure?

It measures what US learners chose to study. Coursera reported that AI security enrollments rose 358% year over year among US learners, against 67% global growth in overall AI enrollments. People are not just taking more AI courses. They are shifting toward the security side of AI, and doing it on their own, often ahead of any employer program.

The figure comes from Coursera's Global Skills Report 2026 announcement, which draws on learning data from more than 300 million Coursera and Udemy learners. The same release reports agentic workflow enrollments up 236% as larger organizations scale up autonomous agents. The global AI security number was even steeper: CIO Dive's coverage of the report put worldwide growth in AI security training enrollment at 665%.

Coursera's own report summary adds a useful piece of context from Deloitte research: 43% of leaders describe their technology infrastructure as highly prepared for AI, but only 20% say the same about their talent. When infrastructure readiness runs more than twice as high as people readiness, the people tend to go looking for the missing skill themselves. That is what the enrollment curve shows.

The assumption worth challenging

The common belief inside smaller companies is that AI security is a technical layer someone else handles. The managed service provider configures the tenant, the vendor promises enterprise-grade protection, and everyone else just uses the tool. Under that model, AI security training is something for the IT person, if there is one.

That model breaks the moment an employee can point an AI tool at company data. The attack paths that matter most in everyday AI use run straight through ordinary users. The OWASP Top 10 for LLM applications ranks prompt injection first, and specifically describes indirect prompt injection, where an AI system reads a website or file containing content that alters its behavior. The person who asked the assistant to summarize that file is the one positioned to notice the output drifting. No firewall rule sees it.

From a security and platform perspective, that is the real reason the enrollment numbers matter. AI security has become an end-user competency in the same way phishing awareness did a decade ago. The controls still matter enormously, but they now depend on the judgment of people who were never trained to think about how an AI tool can be manipulated, what it can reach, and how to shut it off.

Why are security leaders ranking AI above cloud security?

Because AI is creating new exposure faster than existing training covers it. In ISC2's 2026 Security Training Trends research, 47% of security leaders named AI as the most pressing skill their organization is addressing through training, ahead of cloud computing security and security analysis at 44% each and risk assessment at 40%.

The same survey of 995 cybersecurity team leaders found 73% saw their training budgets rise over the past year, and 53% named time and scheduling as the primary barrier to effective training. Even well-funded enterprise security teams, in other words, are struggling to make room for AI security skills. A 40-person company without a dedicated security hire is starting from further back, and has fewer hours to spare.

Cloud security took years to become a standard expectation. AI security is being pulled into that position in a fraction of the time, which is why a training plan written even 18 months ago probably treats it as an optional advanced topic.

Can your team actually stop an AI system that misbehaves?

Many organizations cannot say. ISACA's 2026 AI Pulse Poll of more than 3,400 digital trust professionals found that 56% do not know how long it would take to halt an AI system in response to a security incident, and 39% do not know whether their organization has a documented process for shutting down or overriding AI systems when things go wrong.

Those respondents are auditors, risk managers, and security professionals describing their own employers. If the people whose job is AI oversight are unsure, it is fair to assume the marketing coordinator running an AI-assisted outreach workflow has never been told where the off switch is. The same poll found 90% believe employees are already using AI at work, so the exposure is live even where the response plan is not.

ISACA's reaction tells you how seriously the credentialing world takes the gap. In April 2026 it launched the Advanced in AI Risk (AAIR) certification, joining its recently introduced Advanced in AI Audit and Advanced in AI Security Management credentials. Those are specialist credentials, but they confirm the direction: AI security is now a defined discipline with its own body of knowledge, not a footnote to general IT security.

SignalWhat the data showsWhat it means for a small team
Learner behaviorUS AI security enrollments up 358% year over year, versus 67% global growth in AI enrollments overallYour people may already be learning this on their own, without your context or rules
Security leader priorities47% rank AI as the top training need, ahead of cloud securityAI security belongs in your baseline training, not an advanced elective
Incident readiness56% unsure how fast they could halt an AI system; 39% unsure a shutdown process existsDocument and rehearse the stop procedure before you need it

What should AI security training cover for non-security staff?

Keep it narrow and behavioral. Non-security employees do not need model architecture or red-teaming technique. They need to know what each AI tool can reach, how manipulation looks in practice, what never goes into a prompt, and exactly what to do when an AI system behaves outside its job. Those four skills can be taught in hours and tested directly.

In the environments we support, the training that holds up covers four areas:

  • Access awareness. Which files, mailboxes, and systems a given AI tool or connector can read or act on, and why "it can see everything I can see" is a risk in itself. Our guide to AI connector security goes deeper on what to check before approving one.
  • Manipulation recognition. How indirect prompt injection shows up in a summary, an email draft, or an agent action, and why an output that suddenly asks for credentials or links elsewhere should be treated like a phishing attempt.
  • Data boundaries. What never goes into a general-purpose AI tool. A clear never-paste list does more here than a long policy document.
  • The stop procedure. Who to call, what to disconnect, and how to revoke a tool's access when something goes wrong, rehearsed at least once so it is not invented under pressure.

Notice that the last item is not only a training topic. It is a platform control. Training people to report a misbehaving AI tool is pointless if nobody has defined how to disable its connection, revoke its tokens, or roll back what it changed. That pairing of skill and control is the part most AI security training programs skip, and it is the part ISACA's numbers suggest most organizations are missing.

Why the risk lands harder on smaller teams

A large enterprise that discovers an AI tool leaking data has a security operations team, legal counsel, and an incident playbook. A 60-person firm usually has one internal IT generalist and an outside provider, and a single mistake carries more weight against revenue and client trust. For firms in healthcare, financial services, or professional services, an AI data handling error is also a compliance event, not just a technical one.

The threat side is moving too. Attackers already use AI to sharpen phishing and impersonation, as we covered in how cybercriminals use AI to launch smarter attacks. Employees who understand how AI can be manipulated are better positioned to spot both the attacks aimed at them and the ones routed through their own AI tools. That is the same logic behind building a human firewall, extended to a new class of tool.

How Securafy helps

We approach AI security training from the infrastructure outward. Before anyone sits through a course, we inventory which AI tools and connectors are active in your environment, what permissions they hold, and what data they can reach, then tighten the controls that should not depend on user judgment at all. Our guidance on securing Microsoft Copilot and AI agents before they become shadow IT describes that first step in more detail.

Training then targets the gaps controls cannot close: recognizing manipulation, respecting data boundaries, and executing a documented stop procedure that we build and test with you. If you want a broader view first, our overview of AI security risks and controls for small businesses lays out where to begin, and a short AI readiness assessment shows where your team's AI security literacy stands today.

Where To Go From Here

The enrollment surge says individuals already see AI security as a core skill. Make it part of your baseline training, and pair it with a documented, rehearsed way to stop any AI tool that misbehaves.

If your team is moving faster with AI than your guardrails are, start with structured training rather than another tool. AI University gives your people role-based AI training with security built into the material, not bolted on afterward.

If you would rather talk through your specific environment first, book a strategy call with Securafy and we will walk your current AI usage, exposure, and the fastest path to safe adoption.

Rodney Hall
Rodney Hall

Rodney Hall is the President and COO of Securafy, with 2 decades of experience in IT service management and operations.

He writes about the less glamorous but essential side of IT: support systems, documentation, business continuity, recurring issues, downtime, and the processes that keep client environments running well. His perspective comes from years spent improving how service is delivered, how teams respond, and how small problems are prevented from becoming much larger ones.

Outside of work, Rodney enjoys home improvement projects, woodworking, and dirt bike riding. His personal mission mirrors Securafy’s: helping businesses stay secure, compliant, and ready for whatever comes next.

Writes about: Managed IT, IT operations, service delivery, business continuity, downtime prevention, support processes, operational risk

More from Rodney Hall

Learn AI by building with it

AI University helps teams move beyond AI curiosity through practical lessons, secure workflows, guided experiments, and real projects built for everyday business use.

Explore AI University

Stay current on practical business AI

Get practical updates on AI security, governance, tools, compliance, and implementation without the daily hype cycle.

Join the conversation

Have a question or a different take on this? Add it below.