New 'BluePhishing' scam sees criminals spoof AI agents to get access to corporate environments
Be careful which agents you approve - some are malicious and can bring their friends, too.
A new phishing scam is emerging where cybercriminals are using AI agents to gain unauthorized access to corporate networks, according to security researchers at Proofpoint. The attackers create malicious "agent blueprints" within a company's Microsoft Entra environment, which can then be approved by unsuspecting employees. This allows the malicious AI agents to infiltrate the organization and potentially access sensitive data and resources.
Proofpoint's report reveals how the attack unfolds. The attacker first sets up a malicious blueprint in their own Microsoft environment, configuring the necessary permissions and credentials. They then create a link that directs users to a legitimate-looking Microsoft authorization page. The link, sent via email, social media, or instant messaging, prompts the user to approve the blueprint and its permissions.
When the user clicks the link and approves the blueprint, the malicious AI agent is added to the organization's environment. The higher the target's authority within the company, the greater the potential damage. Ordinary users may only be able to approve certain delegated permissions, depending on the company's policies.
Once the malicious agent is in place, the attacker obtains access tokens for the agent, granting it valid access to various company resources. A single blueprint can support up to 250 agent identities, potentially creating a swarm of malicious bots within the company. The attackers then use these access tokens to exfiltrate sensitive data as they please.
To mitigate the risk of this attack, Proofpoint recommends raising user awareness about the risks associated with granting consent to third-party AI agents, particularly those from untrusted sources. They also advise implementing the principle of least privilege, using controls such as Conditional Access policies for agents and Entra ID protection for agents. Monitoring agent and blueprint events through audit logs can help detect and remove known malicious agents.
Written by urgent.news from TechRadar's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
