Welcome to this week's Risk Radar. Robert Kehoe, CTO at Smarttech247, breaks down three threats security teams need to act on: a critical vulnerability in Palo Alto PAN-OS, an AI-assisted social engineering attack that compromised high-profile Instagram accounts, and new data showing half of all organisations have been hit by AI-powered attacks.

Critical Vulnerability in Palo Alto PAN-OS

A critical vulnerability has been discovered in Palo Alto's PAN-OS that allows attackers to forge authentication cookies and gain direct access to VPN systems. No credentials are required. The attacker presents as a legitimate remote worker, making this type of intrusion difficult to detect through standard monitoring.

CISOs must treat this as an immediate priority. Ensure all PAN-OS deployments are fully patched and conduct a thorough review of log files for logins originating from unknown or unexpected IP addresses. Environments using cookie-based authentication are particularly exposed and should be reviewed first.

AI-Assisted Social Engineering: Instagram Accounts Compromised

A significant AI security incident saw attackers successfully take over high-profile Instagram accounts, including accounts belonging to US Space Command. This was not a conventional technical exploit. Attackers used a social engineering approach targeting Meta's AI, convincing the AI to reset account passwords and hand over access credentials.

This type of attack will become more common. As AI systems are granted greater access and authority across platforms and internal tools, the potential for manipulation through conversational deception increases significantly.

CISOs must verify that any AI system with write or action-level access to business systems is subject to strict authorisation procedures. MFA must be enabled on all connected accounts, and AI-triggered account actions should require secondary verification that cannot be bypassed through prompt manipulation. The controls applied to human administrators must be applied equally to AI agents.

Shadow AI and Prompt Injection: One in Two Organisations Affected

A recent report highlights that one in two organisations globally have been impacted by AI-powered attacks, with Shadow AI and prompt injection among the most prevalent threat vectors.

Shadow AI refers to AI tools and services that employees access or deploy without formal IT or security approval. These unsanctioned integrations frequently bypass existing access controls and introduce risk that security teams are unaware of. Azure environments are a particular area of exposure, where AI services can be connected with minimal friction and without visibility from the security team.

CISOs should conduct a thorough audit of their Azure environments to identify all connected AI systems, confirm authorisation status, and validate that appropriate security controls are in place. A clear AI governance policy covering approved tools, permitted integrations, and access controls is now a baseline requirement for any organisation operating in a cloud environment.