Akamai Firewall for AI claims real-time prompt-injection, jailbreak, adversarial-input, and unsafe-output controls.
Detects and blocks prompt injection, jailbreaks, and adversarial inputs
Public-source research
Vendor evidence
Review what vendors say publicly, the exact quoted source text, the security requirement each statement may support, and what still needs verification.
Research library coverageCounts describe the research workflow, not vendor quality or product effectiveness.View details
Evidence records1280
Source checked863
Needs verification0
Source captured0
No supporting claim found411
Excluded from evidence6
Showing 1–20 of 70 evidence records
Clear all filtersSalt claims real-time protection against abuse of application programming interfaces (APIs) and Model Context Protocol (MCP) interactions used by AI agents.
real-time protection against AI agent abuse
Imperva claims real-time prompt-injection and jailbreak defense for homegrown generative AI applications.
Dynamically detects and neutralizes user inputs intended to manipulate AI behavior
Kong claims standardized PII, authorization, and prompt guardrails for large language model (LLM) and agent applications.
PII sanitization, authorization, prompt guards
Speakeasy claims real-time inspection and enforcement across prompts, responses, and agent actions.
Every prompt, response, and agent action is inspected and enforced in real time.
Harness claims real-time detection and blocking of prompt-injection attacks against running AI applications.
Detect and block prompt injection attacks in real time
Wallarm claims AI payload inspection policies that detect and block AI-agent attacks.
AI payload inspection mitigation controls
Cequence claims real-time guardrails that block prompt injection and business-logic abuse.
block prompt injections and business logic abuse
Upwind claims AI application testing for prompt injection, jailbreaks, unsafe tool bindings, and hallucination-driven data exposure.
Prompt injection and jailbreak testing
Proofpoint claims runtime protection for custom AI models and applications developed or deployed within the enterprise.
protecting custom AI models and applications developed or deployed within the enterprise
Securiti claims controls that mitigate AI risks described by OWASP's large language model (LLM) risk categories and NIST adversarial machine-learning attacks.
Mitigate AI security risks such as OWASP top 10 issues for LLMs and NIST Adversarial Machine Learning Attacks.
Snyk Evo claims continuous security coverage across AI development and AI-native applications.
continuous visibility, governance, testing, and real-time control
ModelOp materials reviewed did not provide a public claim for security controls for custom generative AI applications at runtime.
No quoted source text is recorded for this claim.
LatticeFlow AI claims detection and monitoring of security vulnerabilities, adversarial exploits, data leakage, and other AI risks.
security vulnerabilities and adversarial exploits to hallucinations, bias, data leakage, and compliance gaps
Runlayer claims runtime security inspection before agent actions reach enterprise systems.
Secure runtime execution
Keycard materials reviewed did not provide a public claim for security controls for custom generative AI applications at runtime.
No quoted source text is recorded for this claim.
AWS claims Bedrock Guardrails can block harmful content and use Automated Reasoning checks to reduce hallucinations and data ambiguity.
Bedrock Guardrails can help block up to 88% of harmful content and identify correct model responses with up to 99% accuracy to minimize hallucinations and data ambiguity using Automated Reasoning checks.
Google claims Model Armor filters prompts and responses to protect large language models (LLMs) from malicious or sensitive content exposure or generation.
Model Armor filters both input (prompts) and output (responses) to prevent the LLM from exposure to or generation of malicious or sensitive content.
Microsoft claims Prompt Shields are Foundry guardrail controls for model deployments and agents.
Prompt Shields are part of the Foundry guardrails and controls system
Zscaler claims AI red teaming can conduct vulnerability assessments and simulate attacks on AI systems.
Conduct vulnerability assessments and simulate attacks on your AI systems.