Digital Content Editor, Eve Goode spoke exclusively with Shane Barney, CISO at Keeper Security about how organisations and security leaders should be preparing for AI attacks.
What is the most immediate change security leaders should make to prepare for persistent AI-driven attacks?
The most immediate change is treating every AI agent and automated process as a privileged identity from day one.
Most organisations are deploying agents faster than they extend the identity governance necessary to secure them, and attackers are actively exploiting that gap.
Extending the same controls already applied to human administrators is critical for cybersecurity today.
That means enabling least privilege access, time-bound credentials and full session isolation and visibility for every agent, script and service account.
Organisations cannot afford to wait for a dedicated AI governance framework before acting, because the window for exposure is already open and it is only widening.
Security leaders should create an inventory of all AI agents and automated processes active in their environment, because you cannot govern what you cannot see.
Most organisations vastly underestimate how many AI agents, non-human identities and automated processes are already active inside their environment today.
From there, apply existing access discipline immediately and formalise the broader policy work in parallel.
Which existing security controls are most likely to fail when attacks are operating at AI speed and scale?
Controls built around periodic review cycles are often the first to fail.
Quarterly access reviews, manual log analysis and rule-based alerting were designed for threats measured in days or weeks.
AI-driven attacks compress that timeline to minutes, sometimes seconds, so a security team relying on scheduled checkpoints is structurally incapable of responding in real-time.
Standing privileged access is another critical security control that routinely fails under the duress of attacks at machine speed.
Standing administrator credentials and broad service account permissions were tolerable when a human had to manually exploit them, but today they represent a far more dangerous threat because automated systems can identify and use that access at machine speed.
Detection tools calibrated to known attack signatures also struggle, because AI-generated attacks vary each time they run, evading pattern-based defences.
The common thread across all three is time.
Any control built on the assumption of a human-paced attack, whether that assumption sits in a review cycle, a credential’s lifespan or a detection signature, is being tested by an adversary that no longer operates on human timelines.
How should organisations govern AI agents and non-human identities differently from traditional service accounts?
Traditional service accounts were typically provisioned once, given broad standing access and rarely revisited unless something broke.
AI agents cannot be governed that way because their behaviour is dynamic.
Access needs to be scoped to the specific task an agent is performing, time-bound rather than persistent and reviewed on a cadence that matches how quickly agent permissions change in practice.
Every agent needs a single accountable human owner, something legacy service accounts often lacked entirely.
Session visibility today has to be granular.
Where a service account log might record a login and a handful of actions, an AI agent can execute hundreds of decisions inside one session and security teams need enough detail in that record to reconstruct exactly what happened if something does go wrong.
Third-party AI agents deserve particular scrutiny, since a compromise at the supplier level can provide an attacker a path directly into core systems if access was never properly scoped.
Treat every agent as its own class of privileged identity, with its own owner, its own audit trail and its own defined boundaries.
What role will human security teams play when both attacks and defensive responses become increasingly autonomous?
At machine speed, human review is a bottleneck, not a safeguard.
The role is shifting toward setting the policies that govern autonomous defensive systems, specifically defining access permissions, automatic response triggers, escalation protocols and immediate termination criteria.
Judgement calls around ambiguous incidents, regulatory context and business risk still require a human in the loop.
While machines are effective at applying rules consistently and instantly, they are less effective at weighing context that sits outside those rules.
Future-ready security organisations will shift from passive dashboard monitoring to active governance, defining the operational boundaries for autonomous systems and continuously auditing their compliance.
What should CISOs be measuring now to determine whether their organisation can detect and contain an AI-driven attack quickly enough?
CISOs should measure both detection and containment time, broken down by threat category.
Keeper Security’s 2026 global research found that only 28% of organisations detect that type of misuse within minutes.
But detection is only the first half of the equation.
The real measure is mean time to containment – how quickly a session can be terminated once an attack is identified.
Set category-specific targets: credential compromise and data exfiltration attempts should be contained within five minutes; lateral movement and privilege escalation attempts within ten minutes.
Once something is flagged, does containment require manual intervention or is it automated? That gap determines how much damage actually occurs.
CISOs also must measure coverage, specifically what percentage of AI agents and non-human identities in the environment are fully inventoried, owned and governed under least-privilege policy.
Finally, track credential lifespan across machine identities. Long-lived, standing access is the single biggest predictor of how much damage a compromised agent can do before anyone notices.
Together, those measures give a far more accurate picture of readiness than any AI security policy document.
Shane Barney joined Keeper Security as Chief Information Security Officer (CISO) in May 2025, bringing with him more than two decades of cybersecurity leadership in both the public and private sectors. Prior to joining Keeper, Barney dedicated 20 years to the Department of Homeland Security (DHS), serving within the US Citizenship and Immigration Services (USCIS) as both a contractor and federal employee. He began his career at USCIS in the Office of Security and Integrity (OSI), where he played a pivotal role standing up the agency’s national operations center and building DHS’s first classified system. He also collaborated closely with the intelligence and law enforcement communities to strengthen national security efforts.