Last updated: 7 November 2026. Facts are vendor claims checked on that date; nothing was tested independently.

Cisco AI Defense (formerly Robust Intelligence)

Red teaming as a validation feature of a broad enterprise AI security suite.

Visit Cisco AI Defense (opens in a new tab) Back to comparison

What it tests

Claimed on the vendor's own pages. Nothing here was tested independently.

• Prompt injection
• Agent / tool abuse
• Data leakage
• Bias, safety, harmful content
• Multimodal
• Model supply chain
• 200+ attack techniques and subcategories (data sheet)
• Multimodal attacks in 10 languages
• Supply chain is a separate component

Framework mappings

OWASP LLM Top 10, OWASP Agentic Top 10, MITRE ATLAS, NIST AI RMF, EU AI Act. EU AI Act appears as "compliance readiness" in a Cisco blog, not as a formal mapping in the data sheet.

Offering

• Algorithmic red teaming under "AI Model and Application Validation".
• The data sheet claims 200+ attack techniques and subcategories, multimodal attacks in 10 languages, and mappings to MITRE ATLAS, OWASP (LLM and Agentic) and NIST AI RMF.

Pricing

Not public. Not public. Free tier: none stated. See what AI red teaming costs.

Ownership history

Robust Intelligence was acquired by Cisco. Cisco's blog update notes completion on 2024-09-24 (a later Cisco page says October 2024; this site uses the explicit completion note). Cisco calls Robust Intelligence "foundational" to AI Defense.

Alternatives

Three vendors with a similar offering type and overlapping coverage. Chosen by category and claimed coverage, then alphabetically; no ranking is implied.

• Enkrypt AI

  Automated platform Acquired by Anaconda

  Self-serve, credit-priced red teaming and guardrails with a free tier.

• Promptfoo

  Automated platform Part of OpenAI

  The most widely used open-source LLM red-teaming tool. The paid tiers add dashboards, SSO and continuous monitoring. Buyers should note that it is now owned by a model vendor.

• Giskard

  Automated platform Independent

  Combines LLM quality testing (hallucination, RAG) with automated adversarial probing. It suits teams that want an open-source starting point.