AI in cyber operations: Key lessons from Women in Defense
Sharothi Pikar explores risk and human oversight at the 2026 national conference
What you'll learn
- How the consequences of a task should shape AI autonomy and oversight
- Why faster vulnerability discovery changes what cyber teams need to do next
- Where human judgment remains essential as AI takes on greater responsibility
AI is changing how cyber teams identify vulnerabilities and respond to threats. For defense and intelligence organizations, those capabilities also raise questions about how much responsibility to give machines and where people need to remain involved.
As part of Cybersecurity Awareness Month, these questions bring the focus to the decisions behind secure AI adoption.
At the 2026 Women in Defense National Conference, Sharothi Pikar, Leidos senior director of Cyber Growth, Products and Solutions, discussed these challenges during the panel “The AI Stakes: Who Leads, Who Decides, Who Wins.”
The conversation explored how government and industry can put AI to work while managing the risks that come with it.
Match AI autonomy to the risk of the task
In defense and national security, adopting AI means accounting for security requirements and acquisition processes alongside the potential consequences of an incorrect decision.
Lower-risk applications may allow AI to operate with greater independence. When an action could affect mission outcomes, the need for validation and human oversight increases.
That makes the level of autonomy a decision about risk. Organizations need to understand the task they are assigning to AI and establish safeguards proportionate to what could go wrong.
Close collaboration between the people developing the technology and those using it can inform those decisions.
“We have seen that government and industry each have a critical role in AI innovation,” said Pikar. “The strongest innovation happens when that relationship is collaborative, with continuous feedback between operators and technologists.”
Prepare for faster vulnerability operations
As adversaries adopt greater levels of automation, defense and intelligence organizations face pressure to identify and address vulnerabilities more quickly.
Pikar described vulnerability operations as a growing priority: “From our perspective, the future of cyber will include a greater focus on vulnerability operations that give operators the ability to find, prove and fix vulnerabilities at machine speed."
Parcata™ applies AI to vulnerability analysis and remediation, with human oversight for critical decisions.
The challenge extends beyond finding a flaw. Cyber teams need to determine whether it can be exploited and assess the consequences before deciding how to address it.
This is also the challenge explored in “When AI finds bugs faster than humans can fix them”: how defenders can respond when vulnerability discovery accelerates.
Keep people accountable for consequential decisions
AI can accelerate analysis and, in some cases, action. People remain responsible for the consequences.
A system may identify a vulnerability or recommend a response, but operators need to consider the wider mission context. A technically sound recommendation may still have implications for systems or activities beyond the immediate problem.
Human oversight therefore needs to be built into how AI is used, with clear responsibility for decisions that could affect the mission.
Don’t be afraid of AI – it’s here to stay. It’s our opportunity now to understand it, responsibly integrate it and ensure people remain equipped to make the decisions that matter most.
Sharothi Pikar
Senior Director of Cyber Growth, Products and Solutions, Leidos
Key takeaways
- Set AI autonomy according to the risk and consequences of the task
- Faster vulnerability discovery requires teams to assess and address findings more quickly
- Leidos’ Sharothi Pikar emphasized collaboration between operators and technologists to inform AI innovation
