isco has released new findings from the AI Workforce Consortium’ s latest report, AI Agents and the Impact on

Cybersecurity, highlighting how agentic AI is changing cybersecurity roles and creating new skills requirements.

The research found that 49 % of security leaders consider hands-on experience working with AI agents among the hardest competencies to find in entrylevel cybersecurity candidates. This was followed by technical cybersecurity expertise at 48 % and human-centric professional skills at 45 %.

The findings point to increasing demand for professionals who can combine technical knowledge and practical AI experience with human judgement as AI-enabled systems become more deeply embedded in security operations.

Fady Younes, Managing Director, Cybersecurity, Cisco Middle East, Türkiye, Africa, Caucasus and Central Asia

“ As AI becomes more deeply integrated into cybersecurity operations, organizations need people who can do more than operate technology. Security professionals will increasingly need to understand how to work with AI agents, validate their outputs and apply human judgment when decisions carry risk,” said Fady Younes, Managing Director, Cybersecurity, Cisco Middle East, Türkiye, Africa, Caucasus and Central Asia.“ Developing these capabilities will be essential as organizations prepare for the next phase of AI-enabled security.”

The report also found that cybersecurity work is shifting from manual execution towards strategic orchestration. Security professionals are increasingly expected to establish objectives, guardrails and measures of success, while AI agents execute tasks within those boundaries.

As part of the initiative, Delinea is testing Claude Mythos 5.1 for defensive cybersecurity work, initially applying the technology to its own code responsible for storing, brokering and rotating privileged credentials.

Project Glasswing brings together organisations whose software supports critical infrastructure and where successful cyberattacks could have significant consequences. Delinea is adding the AI model to its existing security review and adversarial testing processes to identify potential weaknesses in its code.

Delinea is using Mythos to examine areas including credential and secrets handling, vaulting logic, key management and cryptographic implementation. Testing also covers session brokering and privileged access paths, as well as policy evaluation and just-in-time authorisation.

Findings will be assessed and remediated through Delinea’ s existing product security processes, with recurring issues incorporated into secure coding standards, pre-merge checks and testing.

“ We’ re treating Mythos like any other privileged identity: its access is scoped, temporary and logged,” said Pierre Mouallem, Chief

Pierre Mouallem, Chief Information Security Officer at Delinea

Information Security Officer at Delinea.“ It runs against our code in an isolated environment with no customer data in scope. Anything it finds enters the same product security and disclosure process as a human-reported vulnerability.”

Delinea will also share validated findings and remediation patterns with the Project Glasswing cohort, while following coordinated vulnerability disclosure processes for issues affecting third-party or open-source components.

WWW. INTELLIGENTCISO. COM / MIDDLE-EAST 9