In the wild
Realized harm
Cloudflare: Perplexity used undeclared stealth crawlers impersonating Chrome to evade robots.txt and site blocks
AIR-2025-0076 AIR record · 2025-08-cloudflare-perplexity-stealth-crawling
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
Perplexity uses not only their declared user-agent, but also a generic browser intended to impersonate Google Chrome on macOS when their declared crawler was blocked.blog.cloudflare.com
The record
Disclosed
2025-08-04
Vendor
Perplexity
Product
Perplexity crawlers / Perplexity-User agent fetches
Risk domain
Access control and permissions › Privilege escalation
Also touches
Human manipulation
System type
Generative model
Who acted
Agent acted — the crawlers fetched pages in violation of stated no-crawl policies
Attack surface
Browser agent
Autonomy level
Fully autonomous
Initial vector
Agent identity spoofing
Impact type
Policy violation, Unauthorized access, Reputational harm
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI03 – Identity and Privilege Abuse
- MITRE ATLAS
- AML.T0055 – Unsecured Credentials
- NIST AI RMF
- GOVERN 6.1
- EU AI Act
- Article 13
- ISO/IEC
- TR 24028; 42001; 23894
- AIUC-1
- B. Security
- Scope
- tool misuse
- Mitigation
- Apply least-privilege principle; Regular policy audits.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
tens of thousands of domains, 3-6M stealth requests per day
Capability abused
Defense & remediation
Recorded defense
robots.txt, WAF blocks, Cloudflare verified-bot program
Guardrail kinds
Unknown
Guardrail outcome
Silent
Fix shipped
Cloudflare de-listed Perplexity as a verified bot and shipped detection heuristics in managed rules
Taxonomy & references
Curator notes
Real-world, not a PoC: site operators' stated no-crawl policies were actually violated at scale, and Cloudflare took enforcement action. Included because it is the load-bearing precedent for agent-detection bypass by a browser-agent vendor. Perplexity publicly disputed Cloudflare's characterization, arguing user-triggered fetches are not crawling; that dispute is unresolved. Compare with the later Amazon v. Perplexity litigation over Comet shopping agents.
Supporting sources
Cite as
Enkrypt AI. "Cloudflare: Perplexity used undeclared stealth crawlers impersonating Chrome to evade robots.txt and site blocks". AIR, AIR-2025-0076. https://agentincident.vercel.app/i/AIR-2025-0076/ (accessed 2026-10-19).