As reported by BleepingComputer, AI-powered phishing campaigns are now generating disposable infrastructure — domains, pages, and lure content — at a pace that fundamentally breaks the economics of traditional blocklist-based defenses. Push Security's analysis deserves serious attention from every security team still anchoring their phishing defense strategy around known-bad indicators.

AI Security Alert: Push Security's analysis deserves serious attention from every security team still anchoring their phishing defense strategy around known-bad indicators.

Why This Is a Structural Problem, Not Just a Scaling One

Blocklists were always a reactive tool. They require a URL or domain to be observed, reported, classified, and distributed — a pipeline that historically took hours to days. Against a moderately skilled attacker sending a campaign from a static domain, that lag was tolerable. Against AI-assisted adversaries spinning up hundreds of single-use phishing pages with legitimately registered domains, valid TLS certificates, and dynamically generated content, blocklists are not merely slow — they are architecturally incapable of keeping up.

The deeper issue is that blocklists defend against identity — known-bad addresses — while modern AI-assisted phishing attacks operate entirely on behavior. The domain is new. The IP is clean. The page is rendered fresh. There is nothing for a blocklist to match against until it's already too late for the targeted user.

Who Is Most Exposed Right Now

Organizations most at risk are those whose phishing defenses lean heavily on:

Why This Is a Structural Problem, Not Just a Scaling One
Email gateway URL filtering based on reputation databases or blocklists
DNS-layer security that flags known malicious domains — useless against newly registered infrastructure
Secure Web Gateways (SWGs) relying primarily on categorization engines that haven't yet seen the domain
Security awareness training alone — AI-generated phishing content is increasingly indistinguishable from legitimate communications, making human detection unreliable as a primary control

Mid-market enterprises are particularly exposed. They often lack the telemetry volume that enterprise-scale threat intelligence platforms rely on for near-real-time blocklist updates, yet face the same adversary TTPs as large targets.

The Case for Technique-Based, Browser-Level Detection

The logical response to identity-evasion attacks is detection that ignores identity entirely. Technique-based detection asks not "Is this domain known bad?" but "Is this page behaving like a phishing page?" — regardless of whether it has ever been seen before.

Browser-level visibility is uniquely positioned to answer that question. At the browser layer, defenders can observe:

  • Real-time credential field rendering and form submission behavior
  • Transparent proxy and reverse-proxy patterns characteristic of adversary-in-the-middle (AiTM) frameworks like Evilginx or Modlishka
  • MFA token harvesting mechanics, including session cookie theft patterns
  • Page structure mimicking known login portals without matching their registered domains
  • Rapid content loading patterns inconsistent with legitimate SaaS applications
Shield53 Perspective: The question defenders should now be asking is not "What bad domains are we blocking?" but "What phishing behaviors can we detect regardless of infrastructure?" That reframe is the strategic shift this threat moment demands.

The AiTM Dimension Makes This Even More Urgent

AI-generated phishing is increasingly paired with adversary-in-the-middle (AiTM) proxy frameworks that defeat traditional MFA entirely by harvesting authenticated session tokens rather than credentials. This means that even if a user completes an MFA challenge, the attacker obtains a valid session. Blocklists are doubly irrelevant here — they can't stop the handshake that's already happened through a clean proxy domain.

This convergence of AI-generated lures, disposable infrastructure, and AiTM session hijacking represents a genuinely new threat tier that legacy detection stacks were not designed to address.

Shield53 Recommendations

Immediate Actions

  • Audit your phishing detection stack: Identify what percentage of your phishing controls are blocklist-dependent. If it's the majority, you have a critical architecture gap.
  • Deploy browser-based security controls: Evaluate browser isolation, enterprise browser platforms, or browser-native security extensions capable of technique-based detection at the point of render.
  • Implement phishing-resistant MFA: Transition from TOTP/push-based MFA to FIDO2/passkey authentication, which cannot be proxied by AiTM frameworks. This is non-negotiable in 2025.
  • Enable Conditional Access with token binding: Use continuous access evaluation (CAE) and device-bound session controls to reduce the value of harvested session tokens.
  • Monitor for AiTM indicators: Look for impossible travel, unfamiliar device sign-ins immediately post-MFA, and session anomalies — signals that a credential or token was harvested even when no known-bad URL was visited.

Strategic Posture

  • Begin treating phishing defense as a behavioral detection problem, not a threat intelligence feed problem. Invest accordingly in tooling that supports that model.
  • Include browser telemetry in your SIEM and SOC workflows — it is rapidly becoming as important as endpoint and network telemetry for identity-based attack detection.
  • Brief your leadership on the MFA bypass reality: the assumption that MFA equals phishing protection is no longer valid against motivated adversaries using modern toolkits.

The blocklist era is not ending — it has ended. The organizations that recognize this now and invest in technique-based, identity-layer-aware phishing defenses will be meaningfully more resilient than those still tuning their DNS sinkholes.