The Great AI Jobs Split: Why Cybersecurity Is Hiring While Tech Giants Keep Cutting

Something weird is happening in the tech job market right now.

While Meta, Google, and Amazon announced another 85,000+ layoffs in late 2024, companies like Palo Alto Networks, CrowdStrike, and Anthropic are desperately hiring AI security specialists. Some are offering $300K+ packages for mid-level roles. We've been tracking this split for six months, and the gap keeps widening.

The numbers tell a stark story. Tech companies eliminated 262,534 positions in 2024 (according to Layoffs.fyi), but AI security job postings increased 87% during the same period. LinkedIn data shows cybersecurity roles mentioning "AI" or "ML" grew faster than any other tech category this year.

Here's what nobody's talking about: it's not just about AI versus non-AI jobs. It's about *which* AI jobs.

The Security Premium Just Got Real

Companies building AI systems need someone to secure them. Badly.

Palo Alto Networks added 3,200 employees this year, focusing heavily on AI-powered threat detection. CrowdStrike's headcount grew 28% year-over-year. Wiz (the cloud security startup) raised $1 billion at a $12 billion valuation and is hiring aggressively across AI security roles.

But here's the thing. These aren't traditional cybersecurity jobs with "AI" slapped on the title. They're fundamentally new roles:

• AI Red Team Specialists (people who try to break AI systems before bad actors do)

• Prompt Injection Defense Engineers (yes, this is a real job now)

• AI Model Security Auditors

• ML Pipeline Security Architects

Anthropic alone has posted 47 security-related positions in the past quarter. OpenAI's security team tripled in size this year. Google's AI Safety division? Also expanding while other departments face cuts.

Meanwhile, in the Layoff Zone

The jobs getting cut follow a pattern.

Software engineers working on stable, mature products. Product managers in established business lines. Data analysts doing work that AI can now automate. Recruiters (ironic, given the security hiring frenzy).

Google cut 12,000+ jobs early in the year, then another wave in Q3. Most affected? Teams working on Assistant, Cloud support functions, and internal tools. The cuts weren't random. They targeted roles where AI could absorb 60-80% of the workload.

Dropbox eliminated 20% of its workforce (528 people) in October, with CEO Drew Houston explicitly saying AI would handle much of what those teams did. Duolingo cut 10% of contractors, replacing them with AI-generated content. Chegg's stock collapsed 99% as ChatGPT destroyed their homework help business.

The data is clear on this one: if your job involves doing the same task repeatedly with predictable inputs, you're in the danger zone. If your job involves securing unpredictable AI systems that nobody fully understands yet? You're suddenly valuable.

Who's Actually Winning Right Now

Let's get specific about companies aggressively hiring:

Security-First AI Companies:

• Anthropic (raised $7.3 billion, hiring across safety/security)

• Scale AI (valued at $13.8 billion, adding security QA roles)

• Hugging Face (expanding trust & safety team by 400%)

Traditional Security Going AI-Native:

• CrowdStrike (Charlotte AI platform launch, 800+ new hires planned)

• Palo Alto Networks (Precision AI initiative, massive hiring push)

• Wiz (fastest growing security company ever, 900+ employees now)

Enterprise AI Security:

• Microsoft (expanding Azure AI security team despite other cuts)

• AWS (Security Lake and AI services team growing 40%)

• Palantir (AI Platform security roles, government contracts driving growth)

And here's a surprise: defense contractors. Lockheed Martin, Northrop Grumman, and Booz Allen Hamilton are all hiring AI security specialists like crazy. Government contracts for AI security are exploding.

The Skills Gap Is Getting Absurd

I've been tracking this for months and the talent shortage keeps getting worse.

There are roughly 4,700 open AI security positions right now (across LinkedIn, Yeah, and company career pages). Most have been open for 60+ days. Companies literally can't find qualified people.

What they're looking for isn't what boot camps teach. You need a weird hybrid skill set:

• Understanding of ML model architectures (transformers, diffusion models)

• Traditional security knowledge (penetration testing, threat modeling)

• Ability to think adversarially about AI systems

• Some programming chops (Python, definitely)

The average salary for an AI Security Engineer in San Francisco just hit $284,000 (Levels.fyi data). That's higher than many senior software engineering roles at the same companies.

But here's what's wild: you don't necessarily need a computer science PhD. Some of the best AI red teamers we've seen came from linguistics, psychology, even creative writing. Because breaking AI systems often requires understanding how to manipulate language and context.

The Jobs Getting Redefined (Not Eliminated)

Not everything's binary (safe versus doomed).

Some roles are morphing into something different:

Software Engineers → Now need to write code that integrates and secures AI systems. The job didn't disappear. It changed. Engineers who can implement guardrails, monitor AI outputs, and prevent model exploitation? Still very employable.

Data Scientists → The ones doing basic analytics are struggling. The ones focused on adversarial ML, model robustness, and AI safety? Getting multiple offers.

Compliance/Governance Roles → Exploding. Every company building AI needs someone who understands EU AI Act, emerging US regulations, and industry standards. These roles didn't exist two years ago.

Technical Writers → Shrinking overall, but technical writers who can document AI security procedures and create incident response playbooks? Actually in demand.

What You Should Actually Do Right Now

Most advice you'll read gets this wrong. They'll tell you to "learn AI" or "get certified." That's too vague and probably too late if you're starting from zero.

Here's the real playbook:

If you're in tech already:

Take our AI job security assessment first. Seriously. You need to know where you actually stand before making moves.

Then pivot toward security-adjacent work in your current role. Volunteer to review AI implementations. Ask questions about model security in meetings. Start a working group on AI safety at your company. Build visible expertise fast.

If you're in cybersecurity:

Learn enough about ML to be dangerous. You don't need to build models. You need to understand how they work well enough to break them. Resources:

• OWASP's LLM Top 10 vulnerabilities (free)

• Anthropic's red teaming guidelines (also free)

• Hands-on practice with prompt injection attacks

Start documenting your experiments publicly (blog, GitHub, LinkedIn). Visibility matters more than perfection.

If you're entry-level or switching careers:

This is actually your best opportunity in years. The field is so new that traditional credentials matter less than demonstrated ability.

Build a portfolio of AI security work:

• Document 10 different ways to break ChatGPT's guardrails

• Audit an open-source AI project for security issues

• Create a framework for testing AI systems

• Write detailed case studies of AI security failures

Companies hiring for these roles care way more about whether you can think adversarially than whether you have a degree.

Specific companies to target right now:

Early-stage security startups (they're desperate and will train you): Protect AI, Robust Intelligence, CalypsoAI, HiddenLayer, Cranium. These companies have more open roles than qualified candidates.

Consulting firms pivoting to AI security: Deloitte, PwC, and Accenture are all building AI security practices and hiring aggressively. Less sexy than startups, but stable and willing to train.

The Timeline You're Working With

This window won't stay open forever.

Right now, companies are so desperate for AI security talent that they'll hire people with adjacent skills and train them up. In 12-18 months? The first wave of university programs will graduate students with formal AI security training. The self-taught advantage narrows considerably.

We're also seeing the first signs of AI security boot camps (still garbage, mostly). Within a year, there'll be a flood of people with surface-level credentials competing for these roles.

And the compensation premium? That'll compress once supply catches up with demand. The $300K packages for mid-level roles won't last.

But the fundamental shift is permanent. AI systems will need securing. That's not a trend, it's infrastructure.

The Real Question Nobody's Asking

Why is AI security growing while AI development is getting automated?

Think about it. AI can help write code. It can analyze data. It can generate content. But AI can't secure itself. Not yet, anyway. Maybe not ever.

Adversarial thinking requires creativity, unpredictability, and a willingness to break things in ways nobody anticipated. That's still fundamentally human work.

Every new AI capability creates 10 new attack vectors. Every model deployed at scale becomes a target. The complexity compounds faster than the tools to manage it.

So while companies cut positions that AI can absorb, they're desperately hiring people to protect against what AI enables.

That's your opportunity. The question is whether you'll take it.

Next steps: Take our assessment to see where you stand. The results are uncomfortably specific about which skills will matter and which won't. Then decide if you're going to adapt or hope this somehow doesn't apply to you.

The data suggests hoping isn't a great strategy.