AI Security Roles Explained

What roles do AI security companies hire?

AI security companies hire people who build the product, people who research how AI systems fail, and people who get the product working inside a customer's business. The last group is the largest and the hardest to define. When we counted 151 open roles at 12 AI security companies on 15 November 2026, 49% were in sales or customer-facing technical work, 28% in engineering and 7% in research. The full count is in our AI Security Hiring Tracker.

The titles aren't standard yet. The same job is called a solutions engineer at one company and an AI architect at another, and one live advert we read used both for the same role. This guide describes what each title usually involves, going by what the adverts ask the person to do.

What does an account executive at an AI security company do?

They own the sale from first conversation to signed contract, selling a technical product to technical buyers in a market that is still being defined.

Day to day that means running enterprise sales cycles, turning interest into paid pilots, managing the proof of concept and working closely with founders, engineering and product. The adverts ask for credibility with technical buyers and the ability to learn a complex product quickly. They also expect the seller to build pipeline without a full SDR team behind them: in our count there were seven closing roles for every SDR or BDR role.

It gets confused with the same title at a mature vendor, where the leads, the playbook and the sales engineer are already in place.

What does a sales engineer or pre-sales solutions engineer do?

They are the technical half of the sale. They work alongside the account executive to understand the customer's environment, show the product and design how it would fit.

The work is technical discovery, demonstrations, proof-of-concept work and solution design, all before the contract is signed. The adverts ask for security depth, enough AI and cloud knowledge to be credible with the customer's engineers, and the ability to explain it to a buyer. You'll see the same job advertised as sales engineer, solutions engineer, solutions architect or AI architect. 17 roles in our count carried one of those titles.

Some companies split pre-sales from post-sales and some expect one person to do both. Ask which before you start a search.

What does a post-sales solutions engineer, customer engineer or technical account manager do?

They make the product work after the sale. The job is deployment, integration and making sure the customer gets enough value to renew.

They onboard the customer, configure and integrate the product in the customer's systems, handle the harder technical problems and feed what they learn back to the product team. The adverts ask for hands-on skill with cloud, containers and networking, and for someone a customer's engineers will trust. Titles include post-sales solutions engineer or architect, customer engineer, customer success architect, technical account manager and deployment engineer.

In AI security these are engineering-capable roles, which is easy to miss when the department is called customer success.

What is a forward-deployed engineer in AI security?

A forward-deployed engineer is a software engineer who builds and ships inside the customer's environment. The role sits with the customer, but the work is engineering.

They design and write the integrations between the product and the customer's systems, turn loosely defined requirements into a working deployment and own the technical outcome end to end. The adverts ask for real software engineering ability, comfort with ambiguity and the confidence to work directly with a customer's engineering and security teams.

The title is still rare: three of the 151 roles in our count used it. It gets confused with sales engineer. The simplest test is who writes production code. A sales engineer demonstrates and scopes, and a forward-deployed engineer builds.

What does an AI red teamer or AI security researcher do?

They attack AI systems to find out how they fail. The work covers prompt injection, jailbreaks, data leakage, misuse of agents and tools, and whether a model can itself be used to carry out cyber attacks.

In practice they design and run adversarial tests against models and agents, build evaluation challenges, research new attack methods and turn the findings into product detections or published research. The adverts ask for a background in offensive security or vulnerability research, with enough machine learning knowledge to understand what is being attacked. Titles include security researcher, cyber researcher, vulnerability researcher and AI red teamer, and at the senior end, head of cyber safety.

The mindset carries over from penetration testing, but the target behaves differently. The same input can produce different outputs, and the weakness is often in how the system reasons.

What does an AI researcher at an AI security company do?

They study what advanced models and agents are capable of, and how to measure and limit the dangerous parts. It is research in the academic sense, applied to security.

They develop methods to test model capabilities, check how well those tests generalise, work on ways to reduce harmful behaviour and often publish. The adverts ask for a strong research record in machine learning and an interest in security and safety.

The security researcher above starts from the attack. The AI researcher starts from the model.

What does an AI security engineer do?

They build the product: the detection and guardrail systems, the machine learning behind them and the platform it all runs on. The adverts are for backend, platform, machine learning and infrastructure engineers, often under titles such as machine learning engineer, software engineer or member of technical staff.

The same title is used inside enterprises for a practitioner who secures that company's own use of AI. One builds a product and the other defends an organisation, so check which you're reading.

Why do these titles get confused?

Because one hire is often asked to cover several of them. In the first edition of The Trace I wrote that forward-deployed engineer, solutions architect and sales engineer can mean quite different jobs from one company to the next. If a brief needs someone to own production engineering, deployment, customer success and the sales process, it's worth checking whether that is a realistic job for one person.

Before starting a search, I'd want to understand three things:

  1. What does this person need to get done in their first six months?
  2. Where will they spend most of their time: building, deploying, supporting customers or helping win business?
  3. Which skills must they bring on day one, and which can the team help them develop?

What do candidates want to know before they join?

For commercial roles, three things kept coming up when we spoke to US security sellers for edition 02 of The Trace: confidence in the product, trust in the person they'd work for, and evidence that other sellers had already succeeded. It was a small group of conversations and we didn't run it as a survey. A hiring team that can explain how a recent customer was won, how the target was set and who will help deliver it will find the role easier to fill.

By Gary Billings, Founder, TechTrace Partners. Based on public job adverts read on 15 November 2026 and on The Trace editions 01 and 02.

AI Security Hiring Tracker, November 2026151 open roles at 12 AI security companies, by function and region