As reported by The Hacker News, Microsoft has disclosed two distinct but equally concerning campaigns — one leveraging AI-generated business email compromise (BEC) at unprecedented scale, and another using passkey-themed lures to hijack Microsoft 365 cloud accounts and exfiltrate data via Microsoft Graph APIs. Both campaigns underscore a hard reality: the security community's enthusiasm for passkeys as a phishing-proof panacea is premature.

Cloud Security Alert: As reported by The Hacker News, Microsoft has disclosed two distinct but equally concerning campaigns — one leveraging AI-generated business email compromise (BEC) at unprecedented scale, and another using passkey-themed lures to hijack Microsoft 365 cloud accounts and exfiltrate data via Microsoft Graph APIs.

The Passkey Illusion

The most alarming element here isn't the BEC campaign — that's a known and growing problem, albeit now supercharged by generative AI. The real wake-up call is the second campaign, active since May 2026, where attackers used passkey-themed social engineering to compromise cloud accounts. Passkeys have been marketed as phishing-resistant authentication. That framing has always been technically accurate but operationally incomplete.

Passkeys resist automated credential harvesting and traditional credential phishing. They do not resist an attacker who can socially engineer a user into registering a new authentication method or approving a malicious enrollment flow on a lookalike domain.

The threat actor playbook described — suspicious sign-ins followed by adding attacker-controlled authentication methods, then high-volume Graph API calls for SharePoint, OneDrive, and mailbox data — is a textbook token-and-session abuse pattern. The passkey theme is simply the latest lure variant. Attackers don't need to defeat the cryptography behind WebAuthn; they need to convince a user to interact with a malicious flow that mimics a legitimate passkey enrollment or verification prompt.

AI-Augmented BEC at Scale

The first campaign — over a million scam emails in a 72-hour window — demonstrates how generative AI has collapsed the cost of personalization. Traditionally, BEC required manual reconnaissance and hand-crafted messaging. Now, attackers can automate executive impersonation with forged email threads, fabricated invoices, and sector-specific language across multiple industries simultaneously.

The use of third-party email delivery infrastructure to bypass sender reputation controls is notable. It means traditional email security gateways that rely heavily on reputation scoring are increasingly insufficient. The layered social engineering — executive impersonation, vendor branding, fabricated supporting conversations — is specifically designed to defeat the kind of skeptical scrutiny that finance teams are trained to apply.

What Defenders Should Actually Be Doing

The default response to these campaigns — "deploy passkeys everywhere" — is necessary but insufficient. Organizations need layered identity defenses that assume users will be social engineered regardless of authentication method.

Shield53 Recommendations

What Defenders Should Actually Be Doing
Monitor authentication method enrollment: Alert on any addition of new MFA factors, especially passkey registration, outside of expected IT-managed workflows. Microsoft's Entra audit logs expose these events — surface them in your SIEM.
Restrict Microsoft Graph API access: Implement Conditional Access policies requiring compliant devices for Graph API endpoints. Monitor for anomalous download volumes, especially bulk SharePoint or OneDrive access from new IPs or unfamiliar geographies.
Enforce managed device requirements: Require device compliance or hybrid Azure AD join for sensitive cloud app access. This prevents token theft from unmanaged endpoints.
Implement payment verification protocols: For finance teams, enforce out-of-band verification for any ACH or wire transfer above a threshold. No email approval should be sufficient — require verbal confirmation via known numbers.
Hunt for the specific IOCs: Look for domains like service-nowinc[.]com and domainlify[.]net in DNS logs, email gateways, and proxy logs. Block and investigate any historical interactions.
Educate on passkey enrollment phishing: Update security awareness training to explicitly warn users that passkey prompts can be spoofed. Users should only enroll passkeys through IT-managed workflows, never from links in emails or external sites.

Broader Implications

The convergence of AI-generated social engineering and passkey-themed phishing represents an inflection point. Attackers are no longer trying to break authentication — they're trying to co-opt it. The security industry must shift from treating MFA enrollment as a one-time onboarding event to treating it as a continuously monitored, high-risk surface area. Every authentication method change is a potential breach signal.

For CISOs, the takeaway is clear: your identity stack is your new perimeter, and it requires the same level of continuous monitoring, anomaly detection, and incident response readiness that your network perimeter demanded a decade ago. Passkeys are a upgrade — not a solution.