US cybersecurity job postings reached their lowest January–September total in four years, despite growing demand for some AI-focused specialists.
CyberSN’s Cybersecurity Jobs Report recorded 235,786 US postings during that period in 2026, down 26.4% from the same months in 2025. The figures measure deduplicated advertisements, not actual hires.
For security leaders, the challenge is balancing AI expertise with the engineering, analysis, and response skills needed to maintain everyday defenses. The report does not establish whether AI adoption caused the hiring decline.
The comparison follows a strong 2025, when January–September postings exceeded the same period in 2024 by about 75,000. However, the latest total also falls below the same period in 2024, so the decline is not solely a comparison with an unusually strong year.
Falling postings also do not mean organizations have all the staff they need. ISC2’s 2024 study estimated a global cybersecurity workforce gap of approximately 4.8 million people, but explicitly stated that this was not a count of current job openings. Staffing needs and funded vacancies are different measures.
Demand surges for AI-related cyber roles
While the total number declined, there are pockets of the cybersecurity market that saw large increases in demand. Postings for AI security engineers and cyber data scientists each increased by more than 200% in January–September 2026 compared with the same period in 2025. AI or machine learning terms appeared in 20% of September postings. Across January–September, however, the share was 14.6%, compared with 15.1% during the same period in 2025.
Growth in these specialist categories contrasts with declines of more than 20% in postings for security engineers, DevSecOps professionals, and security analysts compared with January–September 2025.
For hiring managers, the practical takeaway is to define which AI skills their teams need without overlooking core security responsibilities. Evaluating AI-generated findings, investigating alerts, and maintaining security controls require different capabilities; adding AI expertise should not leave those duties without clear ownership.
Vendors are expanding tools for AI-assisted security work, including Anthropic’s Mythos security-audit offering. OpenAI describes its GPT-5.4-Cyber model as supporting defensive workflows such as analyzing compiled software for vulnerabilities and malware. These capabilities do not, by themselves, establish that an organization can replace a security role or prevent attacks.
What human oversight should involve
Security teams evaluating AI tools should distinguish between generating recommendations and taking action. Summarizing an alert has different consequences from disabling an account, changing access permissions, or deploying a patch.
That distinction should guide controls over what AI agents can actually do. Teams should restrict permissions, identify actions requiring human approval, and retain logs that allow analysts to reconstruct changes and investigate unexpected behavior.
Staffing plans should also account for the work these tools create. Someone must validate findings, maintain integrations, investigate failures, and remain responsible for incident response. A useful evaluation should measure missed threats and incorrect findings alongside any time saved.
Before changing staffing levels, security leaders should establish whether a tool improves a defined workflow and whether the team can detect and recover from its mistakes. Human oversight needs assigned responsibilities and enforceable controls to be effective.
Read more: Before expanding AI across your security operations, consider why effective threat intelligence starts with your existing visibility, detection coverage, and response processes.