AI-enabled attacks highlight supply chain and cloud security risks

In June 2026, the Five Eyes heads of cyber security released a public statement noting that Al models are exceeding expectations, fundamentally transforming both offensive and defensive cyber capabilities. They warned the timeline was months and not years. Four months later, that warning is becoming a reality.

Malicious cyber actors are using artificial intelligence (AI) to accelerate attacks through trusted suppliers and cloud applications. Supply chain and cloud security fundamentals are critical cyber hygiene measures to protect your organisation.

On 21 October 2026, Microsoft reported that a threat actor used an agentic AI framework during a campaign involving compromised ConnectWise application identities. These identities enabled access to customer cloud environments through permissions already granted to the applications. Microsoft observed activity to establish ongoing access and modify privileges in some affected environments. The reporting illustrates how compromised supplier access can expose customers, with AI increasing the speed and scale of malicious activity.

While the techniques that are used by the threat actor are not new, this is an early indication of the scale and speed at which threat actors will be able to perform malicious activities with the assistance of AI. It reinforces our advice that fundamental cyber security controls are the best defence and how this must extend to your suppliers access and integration as well.

Supply chain compromises are becoming more commonplace and the use of AI to rapidly exploit that access to perform at scale compromise is a significant development. Organisations should review third-party access to their systems ensuring they enforce least privilege based on a suppliers function and ensure cloud security controls can detect and contain the misuse of legitimate permissions.

Cyber supply chain risks extend across software, service providers and the organisations supporting them. A supplier’s compromise can create risks for customers that depend on its products or services. Identifying these dependencies, setting security expectations, assessing suppliers and monitoring their security practices throughout the relationship are essential measures of supply chain risk management.

Cloud services require particular attention to access and responsibility. Providers and customers share security responsibilities, but organisations remain accountable for protecting their data. Organisations should understand who manages security configuration, credentials, logging, incident response and recovery for each service they use. ASD’s Cloud computing security for tenants provides guidance on assessing and managing these risks.

In Response to this incident, organisations should:

• Review vendor communications and implement recommended mitigations.
• Disable unrestricted user consent for OAuth applications and require administrator approval for high-risk permissions.
• Review and remove unnecessary permissions granted to third-party applications, enforcing least privilege for application identities.
• Protect application credentials, certificates and service principals with the same rigour applied to privileged administrator accounts.
• Monitor for unexpected application credential additions, new privilege assignments, consent grants, and unusual application activity and promptly investigate any anomalies
• Review existing third-party application integrations and remove applications that are no longer required

Mitigating AI enabled supply chain attacks

Organisations should prioritise the following actions to secure their cyber fundamentals across their supply chains:

• Secure cloud administration and data. Use multi-factor authentication, protect API keys and other credentials, and restrict administrative privileges. Enable security logging and timely alerting. Maintain protected backups and test recovery so that a service disruption or compromise does not prevent restoration of critical data.
• Limit application permissions. Give third-party applications only the access needed for their function and review permission grants regularly. Disable unrestricted user consent. OAuth applications need to be protected with the same security controls as those protecting administrator accounts. Monitor application identities as well as user accounts: the reported campaign used application access without requiring an interactive user sign-in.
• Understand and assess supplier access. Maintain an up-to-date record of providers, connected services, authorised applications and the systems and data they can access. Assess suppliers against the sensitivity of that access and require evidence that they meet security expectations. Include incident notification and response obligations in contracts, and review arrangements as services change.
• Detect changes that could establish ongoing access. Investigate unexpected application credential additions, new privilege assignments and unusual application activity. Monitor service principals and other non-human identities alongside user accounts.
• Maintain visibility and verify remediation. Enable and retain audit logging across cloud services, applications and identity systems. Following a compromise, confirm that unauthorised access, credentials, permissions and persistence mechanisms have been removed, recognising that disabling a compromised application alone may not remove all access established during an intrusion. Validate remediation actions undertaken by suppliers and assess the impact of containment measures on critical business functions.

Prepare for the Future

The significance of this incident is not the novelty of the techniques used, but the potential for AI to accelerate them. Organisations should anticipate faster and more scalable cyber operations.

• Prepare for faster attacks. Rapidly remediate vulnerabilities in internet-facing services, remove unnecessary exposure and verify that security controls work as intended. Exercise incident response plans and use pre-approved, narrowly scoped and reversible automated response actions where appropriate. ASD’s Defending against AI-enabled cyber attacks guidance outlines these measures to help organisations respond to the increased pace of malicious activity.
• Prepare automated Containment actions. Compromises that move at AI speed will need responses that are equivalent, 24/7 365 days a year. Organisations should prepare for automated containment by designing their network and security controls with this in mind, identifying opportunities for automated containment actions to be undertaken with consideration for their business risks and impact, establishing and tightly integrating an orchestration and detection capability. Plan for quick escalation and back out procedures once a human can respond.

Organisations can use the following Five Eyes guidance to review their arrangements:

• Five Eyes cyber security agencies statement
• Information security manual
• Cyber supply chain risk management
• How to manage your security when engaging a managed service provider
• Cloud computing security for tenants
• Defending against AI-enabled cyber attacks: Guidance for government, critical infrastructure and large enterprises