FAANG Cyber Security Jobs: Do You Still Qualify in 2026?
TL;DR — if you've only got 30 seconds
We read 403 of the 743 security postings listed across Apple, Amazon, Google, Meta and Netflix. At every one of the five, some teams now put AI use in the requirements rather than the wishlist.
Application security is the widest door: 15 of the 35 qualifying roles. And 15 of them are base-level, so this is not staff-and-above only.
Search job boards for "agentic" and "AI-assisted", never tool names. Searching for "Claude Code" scored Google 0 of 84; the capability search found 15.
They ask for AI use, not AI knowledge. No certificate satisfies "build and refine Skills, Agents and AI workflows" — only work you have actually done.
The short answer
Yes, you can still get a FAANG cyber security job. But the bar moved, and most people applying haven't noticed.
We read 403 of the 743 security postings listed across Apple, Amazon, Google, Meta and Netflix — a little over half. In every one of the five, we found roles that write AI into the requirements, not the wishlist at the bottom. Apple wants you to "build and refine Skills, Agents and AI workflows, not just operate them." Amazon asks for "2+ years of demonstrated experience utilizing AI/ML technologies for security use cases." Those sit under qualifications, next to the years of experience.
One thing to be straight about before we go further, because it changes how you should read everything here: a job advert tells you what a hiring manager wrote, not what the team actually does. Big companies use standardised, legally-reviewed templates. Plenty of teams are deep into this and haven't updated their job descriptions; others write it in and are barely started. What we can measure is what they now ask for in public, and that is what you're judged against when you apply.
So: what are they asking for?
What changed — AI moved out of the "nice to have" box
This is the part that matters, and it's easy to miss if you skim adverts.
Every job posting has two halves. There's what they require, and there's what they'd like — the "preferred", "desired", "nice to have" section near the bottom. AI language has been drifting into that bottom section for about two years. Nobody had to care. You could ignore it and still get hired.
That's what changed. When we graded these 403 postings we split them by where the AI language sat. We only counted a role if the AI requirement appeared under Responsibilities, Requirements or Qualifications: the things an employer won't hire without. Anything under "preferred" or "nice to have" we deliberately threw out.
There were still 35 roles left across the five companies. That is 35 out of the 403 we read — about one in eleven. This is not "every security job at FAANG now needs AI", and I don't want to leave you with that impression. It's that at every one of the five, some teams have already moved it into the must-have column.
The five of them describe that requirement in completely different language, though, and if you're scanning job boards, that difference will cost you.
What they actually ask for, in their own words
Here's where it gets useful. Read these five side by side and the pattern is obvious — but no single search term would have found all of them.
Apple asks for "AI/agentic fluency" and spells out what that means: "hands-on experience using agentic/LLM-based coding tools. You're able to build and refine Skills, Agents and AI workflows, not just operate them." That last clause is the bar. Using ChatGPT is not the thing being asked for.
Amazon is the most concrete, and the most brutal if you don't have it: "2+ years of demonstrated experience utilizing AI/ML technologies for security use cases, including prompt engineering, LLM integration, or ML pipeline development." Two years. Demonstrated.
Google names its own tool, which nobody else does: "Leverage Gemini and other AI tools to analyze products, identify risks and create remediation strategies." Elsewhere they want engineers to "deploy agentic detection engineering workflows, shifting defenses from 'Detect and Respond' to machine-speed 'Infer and Interrupt'."
Meta names internal tooling you've never heard of: "You'll design autonomous remediation pipelines and AI agents (like Viper and FixieAI) that cut through noise, surface real risk, and patch issues at a scale manual review can't reach."
Netflix is the only one that names the tools you probably already use: "Hands-on use of AI coding assistants / agentic tools (Claude Code, Cursor, MCP) to troubleshoot, automate, and accelerate support work, plus judgment for where GenAI speeds resolution vs. where it doesn't."
Read those five again and notice what they have in common: none of them ask for AI knowledge. They ask for AI use. Apple wants you building workflows, Amazon wants two years of doing it, Netflix wants judgment about when it doesn't help. That's a working practice, not a topic you can revise.
Which raises an awkward question — how do you find these roles in the first place?
How to find them: search "agentic", not "Claude Code"
This one tip is worth the rest of the article, and we learned it by getting it wrong.
When we first went looking for these roles, we searched for tool names — Claude Code, Cursor, GitHub Copilot. Google came back with zero matches out of 84 security postings. Zero. We nearly wrote that up as "Google isn't hiring for this."
It was our search that was broken, not Google's hiring. Google has Gemini. They are never going to name a competitor's product in a job advert. Neither is Amazon, which has Q. Neither is Apple. The only company in the whole set that names Claude Code out loud is Netflix, because Netflix doesn't sell a competing model.
When we searched for the capability instead of the brand — "agentic", "AI-assisted", "LLM", "AI agents" — the numbers moved hard:
If you're filtering job boards by "Claude Code", you are looking at a tiny slice of the market and concluding it's empty. Search "agentic" and "AI-assisted". One word change, roughly eight times the results.
Now you can find them. Which one should you go for?
Which door: AppSec is the way in
Of the 35 FAANG roles that cleared our bar, they break down like this:
Application security is nearly half of them. If you're picking a direction and you want the widest door at these companies, that's it, and it makes sense, because AppSec is where agentic tooling is most obviously useful right now. Reviewing code, triaging findings, chasing vulnerability classes across a huge codebase. That's work an agent genuinely accelerates.
The other number worth knowing: 15 of the 35 are base-level roles. Not staff, not principal. The assumption that FAANG only hires senior people for this is wrong. 12 are senior, 7 are mid-level "II" roles, and one is a manager. The door isn't only open to people who are already there.
Most of these sit in Washington state (15) or are US-wide (9), with a handful in Texas, California and one in the UK. If you're outside the US, that's the honest constraint: this specific set skews heavily American.
So you know the discipline and you can find the adverts. The hard part is proving you can do it.
How to prove it on a CV
This is where most people will struggle, and I don't want to pretend otherwise.
You cannot certify your way into this one. There is no exam that satisfies "build and refine Skills, Agents and AI workflows, not just operate them." Apple is asking whether you have built things. Amazon wants two years of demonstrated use. Netflix wants judgment about when the model is the wrong tool, which you only get from having been let down by one.
What that means practically: the evidence has to be work you've done, not a course you sat.
Show the workflow, not the tool. "Used Claude Code" says nothing. "Built an agent that triages Dependabot alerts against our exploitability criteria, cut review time from 4 hours to 40 minutes" says everything. The second one answers Apple's "build and refine, not just operate."
Name where it failed. Netflix explicitly asks for "judgment for where GenAI speeds resolution vs. where it doesn't." An engineer who can say "we tried agents for X and it was worse, here's why" is demonstrably more useful than one who claims it works everywhere.
Get the two years started now. Amazon's "2+ years of demonstrated experience" is the harshest bar in the set, and it's a clock. It started for some people in 2024. If you begin building this week, you clear that bar in 2028, and the roles that don't ask for two years are the base-level ones we just talked about.
What that looks like in practice: our own receipts
I'll give you ours, because this isn't theoretical for us.
StationX runs on an AI infrastructure platform we built called HAL — 1,400+ skills and commands, 1,281 utility scripts, 20 servers across six cloud providers, 27 categories of persistent memory. It's seven months of compounding work, and it's the execution layer for most of the business. Two things it helped build: JobZone, an AI job-displacement tool scoring 3,500+ jobs across nine countries, in about two weeks. SATs Revision, a maths and English platform with 1,582 questions and nine deterministic validators, in about seven days.
On the security side we run a WordPress bug-bounty pipeline through the same setup, and that's where the honest half of this story lives. The pipeline finds real vulnerabilities — genuine unauthenticated SQL injection, proven with a working proof of concept. It also taught us some expensive lessons, and they're the kind of thing a hiring manager is actually probing for:
We deep-read a finding, logged it, put it on a dashboard — then discovered it was CVE-2025-12752, already patched in the version we'd read. A two-second duplicate check first would have killed it. The rule now is dupe-check before anyone reads code.
We proved a genuine unauthenticated SQL injection in a plugin with 300 installs. Real bug, real proof of concept, worth £0 — below the install threshold that pays. The bug was never the constraint. The target selection was.
Agents are structurally blind to some bug classes. Missing-authorization flaws are 56% of the recent CVEs in our own mirror, and taint scanners cannot see them at all: they hunt dangerous sinks reached by dirty data, not checks that should exist and don't. We only found those by asking a different question, by hand.
That last one is the point. The agent didn't replace the judgment about where to point it — and knowing where your tooling is blind is the thing that separates someone who uses AI from someone who just runs it.
Worth saying plainly: none of this replaces the security fundamentals. Every one of these adverts still wants the Kubernetes, the AWS, the threat modelling, the code review skill. AI use is being added to the requirements, not swapped in for them. That's more work, not less.
And it's being paid for.
What it pays
Of the 35 FAANG roles, 20 publish a salary band. The median tops out at $227,000 and the highest is Netflix at $270,000–$410,000 for a Support Solutions Engineer in Security, Privacy and Assurance, which is the one that names Claude Code by name.
For context, in the wider set of 184 AI-driven security roles we track across 124 employers, the ceiling is higher still: Bridgewater Associates advertised up to $600,000 for a Staff AI Agentic Security Engineer, and Anthropic $320,000–$485,000 for a Staff+ Application Security Engineer.
These are not experimental postings that sit open forever, either. Of 16 roles we tracked from the August waves, 9 had closed within 30 days. They're being filled.
All of which rests on us having read the adverts properly — so here's exactly what we did, and where it's weakest.
How we know
Everything above comes from reading the actual adverts, so here's the method and its limits.
We read 403 security postings across the five companies between July and September 2026, and saved every body to disk so the quotes can be checked rather than trusted. The counts per company:
Meta is the weak one and I'd rather say so than bury it: their careers search renders ten results per page and resists automated collection, so we read 10 of 76. The two Meta roles quoted above are real and captured, but Meta's number is a sample where the others are closer to a census.
We graded every posting by hand against where the AI language sat in the document. Anything under "preferred", "desired" or "nice to have" was excluded, even when the language was strong. That's a deliberately harsh bar and it's why the count is 35 rather than a much bigger, softer number.
Two limits worth repeating. This measures what employers write, not what they do. And it's a sample, not a census — 54% of what was listed, on the days we looked. If we'd read the other 340 postings the count would be higher, not lower, but I can't tell you by how much.
What the sample does establish is direction: at all five companies, for some teams, AI use has already crossed from wish to requirement. That is a floor, not a measurement of the whole.
What to do this week
If you want one of these roles:
- Change your job board search to "agentic" and "AI-assisted" rather than tool names. You'll see roughly eight times as many roles.
- Aim at AppSec if you're choosing a direction — 15 of 35 FAANG openings, the widest door.
- Build one thing and measure it. An agent that does a real piece of security work, with a before-and-after number you can put on a CV.
- Write down where it failed. That's the judgment Netflix is asking for, and almost nobody includes it.
Every listing quoted here, plus more from 124 employers, is on our AI-Driven Cyber Security Jobs page — with the original advert text, the salary, and an archived snapshot where one exists, so you can read the requirements yourself rather than taking our word for the quotes.
Frequently asked questions
Do FAANG cyber security jobs really require AI skills now?
At all five, some of them do. We read 403 of the 743 security postings listed across Apple, Amazon, Google, Meta and Netflix, and found 35 roles where AI use sits in the requirements or responsibilities rather than the "nice to have" list. That's about one in eleven of what we read, so it isn't every security job at these companies — but every one of the five has teams that have already moved it into the must-have column.
Will AI replace cyber security jobs?
Nothing in these adverts suggests replacement. Every posting still asks for the same security fundamentals — cloud, code review, threat modelling, incident response — and adds AI use on top. The requirement is being extended, not swapped out. The risk isn't that the job disappears; it's that the version of it where you do everything by hand stops matching what employers write down.
Which cyber security specialism has the most AI-driven roles at FAANG?
Application security, by a distance. Of the 35 qualifying roles, 15 were AppSec, 9 detection and SOC, 7 cloud, 3 GRC and 1 offensive. That fits where agentic tooling is genuinely useful today: reviewing code, triaging findings and chasing vulnerability classes across a large codebase.
Do I need to be senior to get one of these roles?
No, and this surprised us. Of the 35 roles, 15 were base-level engineer positions, 12 senior, 7 mid-level "II" roles and one manager. The assumption that AI-driven security work at FAANG is staff-and-above only doesn't hold up against what they actually advertise.
How do I search job boards for these roles?
Search for the capability, not the tool. Terms like "agentic", "AI-assisted" and "AI agents" find them; searching for "Claude Code" or "Cursor" does not. Companies with their own AI product — Google with Gemini, Amazon with Q — never name a competitor's tool in a job advert. When we searched by tool name we scored Google at 0 of 84 postings; searching by capability found 15.
What do these roles pay?
Of the 35 FAANG roles, 20 published a salary band, with a median top-of-band around $227,000 and a ceiling of $410,000. Across the wider set of 184 AI-driven security roles we track from 124 employers, the ceiling reaches $600,000 at Bridgewater Associates.
About the Author
Nathan House, Founder & CEO of StationX
Nathan House has 30 years of hands-on cybersecurity experience and is Cambridge-educated, holding CISSP, CISA, CISM, OSCP, CEH, and SABSA. He founded StationX in 1999 — one of the UK’s first cybersecurity companies — and has secured £71 billion in UK mobile banking transactions and the London 2012 Olympics, advising clients including Microsoft, Cisco, BP, Vodafone, and VISA. He authored the world’s most popular cybersecurity course — a #1 Udemy bestseller taken by over 500,000 students — and was named Cyber Security Educator of the Year 2020, AI Security Educator of the Year, and a UK Top 25 Security Influencer 2025. A DEF CON speaker and featured expert on CNN, Fox News, NBC, and the BBC, Nathan leads StationX’s training of more than half a million students worldwide.
