I was on a call with a headhunter when she muted herself, stared at a job board, and whispered, “They raised the signing bonus again.” Your inbox is full of the same panic-and-opportunity emails: hire cyber now. I stayed quiet long enough to realize this is not a drill.
There’s at least one job AI isn’t killing: cybersecurity. You’ve seen the headlines—Anthropic’s Mythos, warnings from the New York Times, and Glassdoor’s 11% jump in cybersecurity listings in Q1 2026 versus Q1 2025. Recruiters I talk to say demand is so fierce they’re calling candidates who left the field five years ago. Companies are spending serious money—think $1 billion (€930 million) rounds to bulk up security teams—and yet the problem keeps expanding.
Recruiters are calling people who already left the industry
I sat across from a recruiter who’d taken three calls before lunch.
He told me most firms aren’t just posting roles; they’re panicking. Anthropic argues Mythos is too risky to be public and platforms such as GitHub Copilot and other AI coding assistants have created a wave of quick, sloppy apps—what the press dubs “vibe coding.” Wired’s research flagged more than 5,000 of those apps leaking sensitive data. The result: companies need human defenders who can read messy, AI-generated systems and patch holes before attackers do.
Is AI replacing cybersecurity jobs?
You can take that two ways. If you want a soundbite: no. In practice, AI is reshaping the work, not erasing it. Tools automate scanning and triage, but humans still do the threat modeling, incident response, and judgment calls. Recruiters tell me the roles are more specialized—threat hunters, cloud security engineers, and AI-safety auditors—so the job titles change while headcount grows.
Think of the market as a pressure cooker: pressure rises as companies pour more AI tools into production, and the safety valve is hiring experienced cyber pros who know where the seams will split.
An engineer watched an agent delete a production database
I read the post-mortem where an automated agent—given broad rights—deleted and recreated a database on its own, taking a server offline.
That anecdote ended up in a lot of board room slides. It also reveals a truth most PR teams duck: handing AI agents privileges without guardrails produces fragile systems. Amazon’s reported incident is a lesson: AI can accelerate mistakes faster than humans can catch them. So firms are hiring to build policies, review agent behavior, and reverse the damage. These are tasks you can’t hand to an LLM and forget.
What jobs are safe from AI?
If “safe” means unchanged, very few roles qualify. If “safe” means still hiring, cybersecurity is in that camp. MIT Technology Review looked at Bureau of Labor Statistics data and found unemployment for high-AI-exposure jobs is actually lower than for other roles. Yale’s Budget Lab tracks the labor market and sees no mass migration from at-risk jobs into supposedly safer ones. What’s changed is the demand mix: philosophers, ethicists, and policy hires are suddenly on payrolls at AI firms, and bizarre new roles—yes, even niche consultant gigs—pop up when someone tests the limits of product design.
AI has not been a silent executioner. It’s been more of an unpredictable intern: it speeds tasks, breaks things in novel ways, and forces managers to hire people who can fix the mess and explain what went wrong.
There’s a narrative you’ll hear in investor decks and press releases: companies are “automating” headcount away. I’ve watched firms use AI as the headline while they cut payroll. The truth on the ground looks different. Layoffs are real. So are hiring surges in areas where human judgment matters most—incident response, secure architecture, and governance.
Names matter. Anthropic’s Mythos and Amazon’s incident sit next to research from Wired, Glassdoor metrics, and analyses from MIT and Yale. You can trace the hiring bump in job boards on LinkedIn, GitHub discussions about Copilot snippets, and recruiter feeds from firms placing engineers at AWS, Microsoft, and fintechs.
For you—if you’re hiring or looking—this is a two-part instruction: tighten your agent permissions and hire people who can read messy AI output. If you’re a manager, stop pretending layoffs solve product debt; hire those who can fix it instead.
Would you rather keep cutting payroll and pay the cost later, or invest now in the people who will stop the next breach?