In conversations with security leaders across the region, Covich describes a consistent theme: AI tools are being deployed widely and quickly across organisations.

But, the underlying security model is struggling to keep up. In his view, this gap is no longer theoretical—it is already surfacing in real-world risk exposure.

His comments align with findings from Proofpoint’s 2026 AI and Human Risk Landscape Report, which quantifies just how widespread AI adoption has become, and how uneven the security response remains.

AI adoption is near-universal—but control is lagging

The report shows that 87% of organisations have AI assistants deployed beyond pilot stages, signalling that AI is already embedded in day-to-day workflows rather than confined to experimentation.

However, only 63% of organisations report having AI-specific security controls in place. Even among those, confidence is low: 52% say they do not fully trust their controls to detect compromise. The result is a widening gap between adoption and assurance.

Perhaps most strikingly, among organisations that do have AI security controls, 50% still reported either a suspicious or confirmed AI-related incident.

As the report frames it, this is not a marginal shortfall in effectiveness—it equates to roughly a coin toss on whether controls are preventing or missing incidents.

A trust problem at the centre of AI risk

The report argues that the core issue is not just tooling, but trust—specifically, trust in data, identity, and communication across increasingly AI-augmented workflows.

As organisations rely on AI systems to act on behalf of employees, customers, and partners, every interaction becomes dependent on the integrity of those underlying signals.

Attackers, the report notes, are increasingly exploiting this trust layer, targeting the spaces where human decision-making, automated systems, and collaboration platforms intersect.

Security struggling to keep pace with adoption

Beyond AI-specific risks, the report highlights broader structural challenges in enterprise security operations:

• Only 33% of organisations say they are fully prepared to investigate AI-related incidents
  
• Just 48% have embedded security considerations from the start of AI adoption
  
• 95% report that managing multiple security tools is a major operational challenge

Together, the findings suggest that complexity itself is becoming a risk multiplier, with fragmented security stacks making it harder to maintain visibility across AI-enabled workflows.

A call for unified security thinking

Covich also points to the value of the report as a tool for reframing the conversation at executive level.

Rather than treating AI risk as a future concern, he argues it should be addressed as a present and measurable exposure that extends beyond email into every collaboration channel used inside modern organisations.

The report concludes that organisations will need to move beyond fragmented tooling and adopt more unified approaches to collaboration security—approaches that can protect interactions across people, systems, suppliers, and AI-driven processes.