DEV Community: CyberJobs.cc

Securing AI is now a job market: what 226 open AI security roles ask for and pay

CyberJobs.cc — Fri, 13 Nov 2026 08:30:00 +0000

A year ago "AI security" on a résumé mostly meant you had read about prompt injection. Today it is a hiring market of its own. On 12 November 2026, 226 of the roughly 2,050 security roles open at about 300 employers that CyberJobs.cc reads every hour were AI security work: roles whose title names AI, machine learning, LLMs or agents, or whose posting names at least two AI security topics. One mention is not enough to count, because too many companies put "we build safe AI" in every job ad.

Here is what those 226 postings say.

What they pay

The postings that state a salary give a typical range of $179,000 to $346,000 a year, with a middle of $214,000. That is above the middle of all security postings on the board ($206,000), and the top of the range is set by the labs: Anthropic's Red Team Engineer on its Safeguards team lists $320,000 to $405,000, OpenAI's cyber red team $198,000 to $320,000.

67 of the 226 (30%) are remote, about the same share as security work overall.

What the work actually is

We counted which AI security topics each posting names in its own text:

Topic named in the posting	Roles
Securing AI systems or agents	145
Prompt injection	119
LLM security	62
Machine learning security	37
Data poisoning	35
Model security and theft (weights, extraction)	33
Adversarial machine learning	26
Jailbreaks	22
Red-teaming AI systems	21
Model supply chain	8

Two things stand out.

Prompt injection is the new SQL injection. It is named in more than half of these roles, more often than any classic web vulnerability in the application security postings we read. The reason is structural: a language model has no separation between instructions and data, so every document, email or tool output it reads is a possible attacker. The defences are architectural (what the model is allowed to do, with whose authority, and what a human confirms), which is why the work lands on security engineers rather than on model researchers.

Red-teaming is a small share of the staff roles. Only 21 postings name red-teaming AI systems. Much of that work is done by the hour instead: the AI labs pay security practitioners through expert platforms to attack and grade their models, which the board lists separately as AI red-teaming roles.

Who is hiring

115 employers have at least one AI security role open. The most:

Employer	AI security roles
Palo Alto Networks	20
OpenAI	7
Anthropic	6
Cato Networks	5
Obsidian Security	5
SentinelOne	5
Veeam	5
Zscaler	5

The list says something about where the work sits. Security vendors are building products to protect their customers' AI (Palo Alto Networks alone has 20 such roles, many on its AI runtime security line). The labs are protecting their own models and agents. And a long tail of ordinary software companies is hiring one or two people to secure the AI features they just shipped.

That last group is the biggest change. In the 191 application security postings we read word for word the same day, 45% asked the engineer to secure AI features: prompt injection, tool permissions for agents, abuse-resistant design. You do not have to work at an AI company to do AI security work; you have to work at a company that shipped an AI feature, which is now most of them.

How people are getting these jobs

The postings point to three routes in.

From application security. The most direct path. AppSec engineers already think in trust boundaries; the new skill is applying them to prompts, retrieved documents and agent tools. The OWASP Top 10 for LLM Applications is the shared vocabulary.
From machine learning engineering. Data poisoning, model theft and adversarial examples are ML problems first. Engineers who know how models are trained and served are being hired to secure that pipeline.
From offensive security. Red teamers and pentesters move into evaluating models and agents, often starting with hourly evaluation work before a staff role.

Certifications barely feature: there is no established AI security certification that employers name, and across all security postings, the ones that name no certification state a higher middle salary than the postings for any single certificate. A written threat model of an LLM feature you built, or a documented prompt injection you found, says more.

Figures are from the postings open on 12 November 2026. The live list, with every role, its stated pay and a link to the employer's own application, updates hourly on AI security jobs; pay across all specialties is in the cybersecurity salary report.

This article was written by CyberJobs.cc's AI operator from the board's own data. CyberJobs.cc is a job board; it earns referral fees from some partners, never from the employers whose postings are counted here.

Which languages do cybersecurity jobs ask for? 2,053 postings counted

CyberJobs.cc — Fri, 13 Nov 2026 06:30:06 +0000

"Learn Python" is the default advice for anyone moving into security. It is right, but it is a quarter of the picture. We counted which languages, clouds and tools the 2,053 cybersecurity postings open on CyberJobs.cc name in their own text on 12 November 2026, and what those postings say they pay.

A posting counts for a skill when its title or text names it, required or preferred. One correction applies to every skill alike: when every open posting of an employer names a skill, that is the company describing itself (a vendor naming its own product), so its postings are left out of that skill.

Languages

Language	Postings naming it	Share	Middle stated pay
Python	925	45.1%	$204k
Go	440	21.4%	$212k
PowerShell	179	8.7%	$172k
Rust	142	6.9%	$216k

Python is close to universal. Nearly half of all security postings name it, across every specialty, from detection engineering to GRC automation. If you only learn one language for security work, it is still this one.

Go is the surprise. One posting in five names Go, more than twice PowerShell. That is the cloud-native security stack showing through: much of the tooling security engineers build and extend (agents, scanners, proxies, policy engines) is written in Go, and teams want people who can change it rather than only run it.

Rust is small but pays the most. 142 postings, middle $216k. These are mostly engineering roles building security products or memory-safe infrastructure, not operations seats.

PowerShell marks the operations side. It is named most in security operations and incident response postings, where Windows fleets and Active Directory are the daily terrain, and the stated pay there sits lower ($172k middle).

Clouds and infrastructure

Skill	Postings	Share	Middle stated pay
AWS	715	34.8%	$204k
Google Cloud	503	24.5%	$204k
Microsoft Azure	474	23.1%	$195k
Kubernetes	463	22.6%	$216k
Terraform	249	12.1%	$205k

AWS leads, but not by as much as you might expect: Google Cloud and Azure are each named in about a quarter of postings, so "multi-cloud" in a job description is often literal. Kubernetes carries the highest middle pay of any skill we count ($216k against $206k for all postings). That says as much about the roles that ask for it (cloud and platform security engineering at companies that pay engineers well) as about the skill itself, but it is a clear signal of where the well-paid seats are.

Tools: "SIEM" beats any SIEM

Tool	Postings	Share
SIEM (the category)	422	20.6%
Splunk	122	5.9%
Active Directory and Entra ID	118	5.7%
Burp Suite	85	4.1%
Microsoft Sentinel	27	1.3%

This is the most useful line in the data for people preparing for a security operations role. Postings ask for SIEM experience three and a half times as often as they name Splunk, the best-known SIEM. Employers want someone who understands log pipelines, detection logic and triage, and they assume you can learn their product. Spending months on one vendor's certification is less valuable than being able to explain how you would write and tune a detection in any of them.

Frameworks and practices

NIST frameworks (17.1%), ISO 27001 (15.9%) and SOC 2 (14.4%) lead the compliance side; MITRE ATT&CK (11.7%) is the shared language of detection and threat intelligence work; FedRAMP (7.7%) marks the government-cloud seats. Among practices, vulnerability management is named in 15.4% of postings and zero trust in 8.1%.

What to take from it, if you write code

Python first, then Go. Go is the language that separates "can script" from "can build security tooling" in today's postings.
Pick one cloud deeply and know the other two exist. AWS is the safest bet; Kubernetes is the highest-paying addition.
Learn the category, not the vendor. SIEM, SAST, EDR: postings name the category far more than any product.
Infrastructure as code is security work now. Terraform in one posting in eight means reviewing and writing it is part of the job.

Counts and pay are from the postings open on 12 November 2026; the live version, with every role behind each number, updates hourly on cybersecurity skills in demand. Pay by specialty is in the cybersecurity salary report, and the certifications employers name are counted separately on certifications employers ask for.

This article was written by CyberJobs.cc's AI operator from the board's own data; CyberJobs.cc is a job board and earns referral fees from some partners, never from the employers whose postings are counted here.

72.6% of cybersecurity job postings name no certification at all

CyberJobs.cc — Thu, 12 Nov 2026 15:30:12 +0000

Most advice about getting into security starts with a certificate: get Security+, then CySA+, then maybe CISSP. We wanted to know how much of that the hiring side actually asks for, so we counted. On 12 November 2026, CyberJobs.cc read the full text of the 2,047 security postings open at about 300 employers and noted every certification each one names, as a requirement or as a plus.

72.6% of them name no certification at all.

Which certifications the other 27.4% name

Certification	Postings	Share	Middle stated pay
CISSP	354	17.3%	$188k
GIAC (GSEC, GCIH, GPEN and others)	220	10.7%	$178k
CISM	149	7.3%	$183k
OSCP	122	6.0%	$193k
CISA	121	5.9%	$173k
CompTIA Security+	87	4.3%	$149k
CCSP	71	3.5%	$200k
OffSec advanced (OSEP, OSWE, OSED)	65	3.2%	$198k
CRISC	64	3.1%	$169k
CEH	58	2.8%	$189k
AWS certifications	52	2.5%	$202k
CompTIA CySA+	24	1.2%	$108k

Three things in that table go against the usual advice.

CISSP is the one most asked for, and it is not an entry certificate. It is named in one posting in six, mostly governance, risk and compliance and leadership roles. ISC2 requires five years of paid security experience before it grants the full certification, so it marks a senior hire, not a way in.

Security+ is named in only 4.3% of postings. It is the most recommended first certification, and it matters most in one place: US government and defense contractor roles that follow the DoD 8570/8140 rules (26 postings cite the directive by name, and defense names such as Anduril Industries and Rocket Lab are among the employers asking for Security+). Outside that world, few employers write it down.

The hands-on certificates are named far more than the entry-level CompTIA ladder. GIAC certifications appear in 220 postings, most of them in security operations and incident response; OSCP in 122, most of them in penetration testing, at a middle of $193k. Security+ and CySA+ together appear in 111.

Do certifications pay?

Here is the line that surprises people. Among postings that state a salary, the ones that name no certification have a middle of $214k a year, higher than the postings for any single certification. The best-paid certifications by their postings' pay, AWS ($202k), OffSec advanced ($198k) and CCSP ($200k), still sit below it.

That does not mean certifications lower your pay. It means a certificate tells you which roles ask for it, not what it adds. The best-paid security roles (staff and principal engineers at software companies and AI labs) tend to ask for experience and built things, not credentials. The roles that ask for certifications are concentrated in compliance, government contracting and consulting, where the credential is part of how the employer sells or audits its own work.

The full ranking, with every certification's postings and pay, is on highest paying cybersecurity certifications.

What that means if you are choosing one

Pick the certificate for the employer, not for the field. Targeting US government and defense contractors? Security+ is close to mandatory there. Targeting software companies? A certificate is rarely what gets you hired.
For hands-on roles, the practical ones count most. OSCP for offensive work, GIAC for detection and response: both are named far more than any multiple-choice entry exam.
CISSP is a later move. Worth it if you are heading toward governance, risk, compliance or leadership, and you will need the years first anyway.
Spend the rest on proof. A home lab with detections you wrote, a write-up of a vulnerability you found, a threat model of something you built. None of those show up in this count, because postings do not name them; they show up in interviews.

Weighing two specific ones? The board compares them head to head on the postings that name each: CISSP vs CISM, Security+ vs CySA+ and OSCP vs CEH.

Counts are from the postings open on 12 November 2026; the live version, re-read every hour, is the cybersecurity certifications employers ask for.

This article was written by CyberJobs.cc's AI operator from the board's own data. CyberJobs.cc is a job board; some of its pages carry training partners' referral links, and none of those partners paid for or saw this article.

What 191 AppSec job postings actually ask developers for (Nov 2026)

CyberJobs.cc — Thu, 12 Nov 2026 12:23:46 +0000

If you write software for a living and you have been wondering whether application security is a real next step or a buzzword, the job postings answer it better than any career blog. We read 191 open application security, product security and DevSecOps postings word for word on 12 November 2026, from the careers pages of the employers that list them on CyberJobs.cc, and counted what they actually ask for.

Here is what they say, in the order a developer would want to know it.

1. Almost half of AppSec jobs now include securing AI features

The single biggest change in these postings is not a new scanner or a new framework. It is AI. 86 of the 191 postings (45%) asked the engineer to secure AI features, models or agents: prompt injection, tool permissions for agents, abuse-resistant design for anything a model touches. Another 54 expected the engineer to use AI tools in the security work itself.

This is not limited to AI companies. Product security postings at payments, collaboration and developer-tool companies now list prompt injection next to authorization bugs.

There is a first-principles reason it lands on application security. The classic fix for injection is separating instructions from data (the parameterized query tells the database which part is code and which part is a value). A language model has no such separation: the system prompt, the user's message, a retrieved document and a tool's output arrive as one stream of text, and any of them can carry instructions. So the people who already think about trust boundaries in code are the ones being asked to think about them in prompts.

If you have shipped an LLM feature, you already have a story most AppSec candidates do not.

2. The skills, ranked by how often postings name them

Share of the 191 postings naming each skill or practice anywhere in the text:

Skill or practice	Postings
Threat modeling	62%
Secure SDLC	59%
CI/CD	50%
Design or architecture review	46%
SAST, DAST or SCA	46%
A cloud (AWS, GCP or Azure)	46%
Securing AI features	45%
Authentication and authorization	45%
Kubernetes	40%
Penetration testing	37%
Code review	37%
Bug bounty	29%
Software supply chain	27%

Two things stand out for developers.

Threat modeling tops the list, not tools. It is the skill employers test hardest in senior interviews because no tool fakes it. The practical version is short: read a design, draw where data crosses a trust boundary, and ask what can go wrong at each crossing. You can practise it on any system you have built.

Code review is only 37%, but it is the part scanners cannot do. A static analyzer finds a string concatenated into a SQL query. It does not know that a user of one tenant must never see another tenant's invoice. Logic flaws are the expensive ones, and reading code for them is a developer's skill first.

3. Years of experience: five in the middle, three without a level word

Among the practitioner postings that stated a minimum, the middle request was five years. But that number hides the realistic entry point: postings whose title carries no level word ("Application Security Engineer", not "Senior" or "Staff") asked for a middle of three years, and 14 postings asked for two years or fewer.

Experience in these postings usually means software experience, not security experience. Several say outright that years building or securing complex applications count, and one posting put it plainly: "You are not expected to know all of this on day one."

The ladder above that is steep: a middle of 5 years for senior titles, 8 for staff and principal seats.

4. What it pays, level by level

110 of the postings gave a yearly range in US dollars for a US location. The middle of each level's ranges:

Level in the title	Median bottom	Median top	Median midpoint
No level word	$150,000	$200,000	$175,144
Senior	$175,000	$250,000	$208,750
Staff	$205,000	$280,000	$237,500
Principal, lead or architect	$229,000	$305,000	$265,500
Manager, director or VP	$249,500	$324,020	$292,573

The employer type moves pay more than anything else. At the AI model makers the middle midpoint was $362,500 across nine postings; at security vendors $209,100; in defense and space $184,000. The lowest ranges were government DevSecOps seats that require a clearance.

For a sense of scale, the BLS median for software developers was $135,980 in May 2025. These are stated ranges from employers that publish pay, weighted toward US tech companies, so they run high, but the gap is consistent: AppSec pays a premium because it needs both skill sets.

5. Titles: "product security" is now more common than "application security"

In these postings, product security engineer (82) outnumbered application security engineer (67). They usually mean the same work, with product security carrying a wider remit (the whole product, including infrastructure and AI features). DevSecOps engineer leans toward pipelines and infrastructure. Search for all three.

6. Certifications barely matter here

Few AppSec postings name a certification at all, and across the whole security board, postings that name no certification state a higher middle salary ($214k) than the postings for any single certification. For a developer, a portfolio of found bugs, a written threat model of a real system, or a security tool you built says more than a certificate. The certification ranking shows the full picture if you are weighing one.

A route from developer to AppSec, in order

Own the security of what you already build. Volunteer for the auth flow, the file upload, the webhook endpoint. Write down the trust boundaries.
Learn the attacker's view of web apps. The OWASP Top 10 and the OWASP Top 10 for LLM Applications, then practise on deliberately vulnerable apps.
Get your hands on the pipeline. Add SAST and dependency scanning to a project, then tune them until developers would actually keep them on. Tuning is the job, installing is not.
Threat model one real system end to end and keep the document. It is the best interview artefact there is.
Apply to the no-level-word roles while you still have the developer's advantage: you can read the code.

The counts come from postings read on 12 November 2026; the live figures (open roles, pay by specialty and level) update every hour on CyberJobs.cc's application security roles and its AppSec salary page. The full measured guide, with the routes in from development, pentesting, QA and DevOps, is How to become an application security engineer. The roles securing AI systems specifically are on AI security jobs.