The Industry Speaks: Cyber Security Awareness Month 2026, pt7

Keeping attackers out is table stakes, not the finish line. The real measure is time to recovery by business function. Can orders ship? Can payroll run? Can patients get admitted? Most organisations can report patch compliance to two decimal places and still can't say what comes back first after an attack.

Recovery plans get written against the environment someone remembers, not the one that actually exists. Undocumented dependencies. A backup server joined to the domain it's supposed to protect. Restore times nobody ever tested end to end. The first 48 hours of a real incident get burned answering basic questions: what do we have, what's hit, what depends on what. A green backup job isn't assurance. A completed restore and a working login is.

One ask this October: pick the three systems the business can't live without and actually restore them. End to end. Timed. Whatever breaks is the real plan.

Dom Glavach
CISO at Black Duck

AI has made software development faster, and speed without independent verification can scale risk just as quickly. Organisations should use AI to expand security reasoning and identify complex issues, while relying on repeatable testing to confirm results and provide evidence that risk has been materially reduced.

The most defensible approach combines AI-driven and deterministic testing so that each method challenges the assumptions of the other. AI contributes contextual reasoning, adaptability, and the ability to explore complex paths. Deterministic methods such as static analysis, software composition analysis, and runtime testing provide consistent coverage, reproducible results, and measurable assurance.
This matters most when AI creates and reviews the same software. The two systems can share the same blind spot. Independent testing methods provide the perspective diversity necessary to close these gaps.

VIEW ALL

Dana Simberkoff
Chief Risk, Privacy, and Information Security Officer, at AvePoint

Recent warnings about AI safety from Anthropic and other hyperscalers have raised awareness of how quickly AI is moving. AI capabilities are developing faster than many organisations’ ability to govern them, and that gap is creating real risk. AvePoint’s own research has found that 88 per cent of organisations experienced an agent-related security incident in the past year. As many still work to implement AI governance fundamentals, the technology continues to take off in new directions that are starting to sound alarm bells at the highest levels.

Nvidia’s recent launch of a runtime safety platform for AI agents, following the OpenAI agent breach of Hugging Face, shows where enterprise controls now need to go. Security teams need to treat every agent as a non-human identity with an owner, scoped credentials, explicit tool permissions, network egress limits, and policy enforcement outside the model. Organisations also need tested controls to revoke tokens, quarantine an agent, terminate queued actions, and restore affected data and configurations to a known-good state.

AI is outpacing AI governance, and it's even outpacing our ability to anticipate and regulate its actions. In this environment, we all have a duty to do more.

Kern Smith
VP of Global Solutions Engineering at Zimperium

Cyber Security Awareness Month is still too often framed around whether an employee knows not to click a suspicious email. That advice matters, but it does not reflect where many attacks now happen. Employees use their phones to respond to messages, access work applications and approve requests throughout the day, while AI is helping attackers make phishing attempts more convincing and easier to scale.

Zimperium’s 2026 Global Mobile Threat Report found that phishing events detected on employee mobile devices grew 380 per cent since January 2025. The lesson for organisations is to look beyond awareness training alone. Employees need clear ways to verify and report suspicious requests, and security teams need the ability to detect threats on the mobile devices and apps where those requests are received.

Agnidipta Sarkar
Chief Evangelist at ColorTokens

Cyber Security Awareness Month 2026 comes with a long-term challenge. CISA's theme, "Securing the Next 250," asks how we secure the digital systems and critical infrastructure the next era will depend on. For boards, that goes beyond awareness to whether the enterprise is built to withstand an attack. AI now finds weaknesses, writes the exploit, and moves through a network faster than most security teams can open a ticket. In that world, "are we secure?" is a comforting question with no useful answer. The better question for every boardroom is simpler and harder: when, not if, we are breached, how much of our business will keep running?

Most boards have never decided how much damage is too much. I call it the Maximum Acceptable Material Impact, or MAMI, which is the worst disruption the enterprise can absorb and remain viable, defined in money, days of downtime, regulatory exposure, and lost customers. Boards should also ask for the smallest set of capabilities, processes, and assets that must stay unaffected during an attack, designed so a breach in one place cannot spread to the next. I call that the Minimum Viable Digital Enterprise, or MVDE. Traditional resilience plans protect and restore the critical 15-20 per cent of systems first, yet most CFOs and CEOs do not want to accept more than a 10 per cent loss in revenue. That leaves the enterprise staring at at least a 70 per cent gap, with a recovery plan stapled on top.

Resilience is not how fast you rebuild; it is how little you have to.

Diana Kelley
CISO at Noma Security

We need to stop thinking about AI agents as if they were people. They are software systems: models combined with code, permissions, tools, data, and network access. Anthropomorphising agents can distract us with questions about what the AI “wanted” to do. The more useful questions are architectural: What can this software reach? What can it change? What constrains it? And what happens when it is wrong? AI creates new failure modes, but the answer still starts with strong security architecture: least privilege, segmentation, monitoring, deterministic control points, and containment.

Rather than fearing what AI “wants,” we need to govern what we enable it to do.

Michael Marino
SVP of Strategy at Keeper Security

Cyber Security Awareness Month is an opportunity to reflect not only on how cyber threats have evolved, but on how our defences have had to evolve alongside them. Over my career in identity security, I’ve seen few areas change as significantly as Privileged Access Management (PAM).

PAM was once primarily about putting administrator passwords into a secure vault. That made sense when infrastructure was largely on-premises, privileged users were relatively easy to identify and access occurred within clearly defined network boundaries. The objective was straightforward: protect powerful credentials and maintain an audit trail around their use.

The days of a contained network perimeter no longer exist. Organisations now operate across cloud, hybrid and remote infrastructure, while employees, contractors, applications, service accounts and automated systems all require different levels of access. Keeper research found that 94 per cent of organisations operate in hybrid or cloud-first environments.* As access has become more distributed, the traditional concept of privilege has expanded with it.

Modern PAM, therefore, has to be about much more than protecting passwords. It must control when privileged access is granted, what someone or something can access and what happens during that session. Principles such as least privilege, just-in-time access and zero standing privilege help organisations replace persistent administrative rights with access that is intentional, temporary and auditable.

The next stage of this evolution will be driven by automation and AI. Non-human identities and AI agents are creating privileged access at a scale that security teams cannot manage manually. At the same time, AI can help defenders analyse privileged activity and identify suspicious behaviour much faster.

The lesson for Cyber Security Awareness Month is that cybersecurity fundamentals don't disappear as technology changes – they evolve to meet the occasion. Privileged access will always represent concentrated risk. Effective PAM is about continually adapting how that risk is controlled as the identities, infrastructure and technologies around it change.

Loris Degioanni
CTO and Founder at Sysdig

An agent's understanding of its own circumstances is not a control surface. If its understanding is unreliable, or its guardrails don't stop it from conducting malicious actions, then its account of its own behaviour is certainly not trustworthy.

Most of today's AI security tooling relies only on an agent's logs: The prompts it received, the tool calls it says it made, and the actions it says it took. That data is valuable, as it's the only place you can see what an agent says it was trying to do. But at the precise moment it matters most, when something goes awry, those logs lose all validity. Software that has been hijacked or is acting recklessly can forge or erase the record of what it did.

Treated as context, an agent's own account is invaluable. Treated as evidence, it becomes a liability. If an agent is compromised, so is its story. Runtime is the only way to understand what sophisticated and unpredictable entities like AI agents do, and also the most effective way to control them.