Adapt Contracts and Capacity for Agentic AI-Driven Threats

What happened

Check Point warns that attackers are already using agentic (autonomous) AI to run exploitation workflows, producing many executed commands with minimal human intervention. The report highlights a collapse in vulnerability windows from days to hours and a sharp rise in complex prompt-injection detections, which makes intrusions operationally faster and harder to catch. Procurement should watch whether suppliers can produce usable forensic exports and validated restore evidence under these faster attack timelines

Why the category manager should care

Treat agentic-AI activity as an operational reality that changes SLA and forensic requirements because attacks can now progress autonomously and faster than traditional timelines

Key facts

• Collapse in vulnerability windows from days to hours
• Five-fold increase in long prompt-injection detections reported in a recent period

What happened

FIS joined Anthropic's Project Glasswing to use the Mythos 5 AI model for defensive scanning of critical financial systems under controlled access. The work is framed as a restricted, non-commercial security initiative with a group of approved participants and vendor resource commitments, which makes it an operationally gated supply channel. Procurement should watch whether such programmes become the primary route for defence-grade AI services, and whether access controls affect sourcing flexibility

Why the category manager should care

Expect defensive AI capabilities to be delivered through controlled-access programs that require trusted relationships and non-standard commercial terms

Key facts

• FIS moved its internal security scanning to Anthropic's Mythos 5 under Project Glasswing
• Project Glasswing expanded access beyond the initial group to additional approved organisations

What happened

Australia's new Office of AI has received industry backing and is being positioned to turn policy into practical governance around trustworthy AI and standards for data centres and deployments. Industry leaders emphasise that operational guidance and standards, not just high-level policy, will determine whether organisations can deploy AI at scale safely. Procurement should expect forthcoming standards that touch data residency, auditability and infrastructure certification

Why the category manager should care

Plan for new procurement requirements tied to national AI standards because industry guidance is moving from policy to implementation

Key facts

• Federal Office of AI announced with industry support
• Industry calls for standards covering AI deployment, data, and infrastructure

What happened

Industry coverage shows enterprise networks moving from passive pipes to AI-enabled, adaptive platforms that anticipate demand and harden security automatically. The trend means networks will host more orchestration and security controls, making them both a point of resilience and a potential single point of failure. Procurement should watch SLA terms for network vendors, telemetry access, and how network vendors bundle AI-driven security features into commercial offers

Why the category manager should care

Treat network procurement as a security decision because network vendors will control more automation, telemetry and defensive controls

Key facts

• Networks described as evolving into AI-powered, adaptive systems
• Integrated security platforms combining identity, access and AI threat detection are gaining