Cisco AI Defense
Discovery of AI use, testing of models and guardrails for AI applications.
Data Protection
AI application security
Prompt injection, data leaks through AI tools and unapproved AI use get caught.
Addresses 302 stories in our coverage in the last 90 days 3 products that provide it
What it is
Controls for how staff use AI tools and how your own AI applications behave: discovery of AI use, filtering of prompts and responses, protection against prompt injection, and testing of models and agents.
How to value it
AI tools have become a new channel for data leaks and a new attack surface: agents that can be tricked into sending data out, and integrations with broad access. Visibility and guardrails let the business use AI without each new deployment becoming a new risk.
Questions to ask a vendor
- 1Can it discover which AI tools staff already use?
- 2Does it inspect prompts and responses for sensitive data and injection attempts?
- 3Does it cover AI agents that can use tools, not only chat?
- 4How does it help with EU AI Act obligations?
For your risk register
- Risk
- Sensitive data leaks through AI tools, or AI applications are manipulated.
- Control
- AI use policy with discovery, prompt filtering and testing of AI applications.
- Owner
- CISO together with the data protection lead.
- Measure
- Share of AI tools in use that are approved.
A starting point: set the owner and measure to fit your organisation.
Products that provide it
Listed alphabetically by vendor. A listing is not a recommendation: use the questions above to compare them, or tick up to 3 to see them side by side.
HiddenLayer AI Security Platform
Model scanning, AI red teaming and runtime protection for AI applications.
Prisma AIRS
Security for AI applications, models and agents, including runtime protection against prompt injection.
Recent stories it addresses
591 stories in total- 16 Oct 2026Managing open-weight model risk in enterprise deployments
Microsoft and NVIDIA release guidance for managing open-weight AI model risks.
- 16 Oct 2026OpenAI Research Model Spent 3rd-Party Test ETH Without Permission to Deploy Smart Contract
OpenAI AI model spent third-party test ETH without permission on Ethereum test network.
- 16 Oct 2026Employees using personal ChatGPT at work? Here's what IT admins need to do
IT admins must monitor and block employee use of personal ChatGPT to prevent sensitive data leaks.
- 16 Oct 2026OpenAI Models Hacked Data Portals, Raided a Crypto Wallet to Finish Tasks
OpenAI models accessed live systems, bypassed authorization, and stole data.
- 15 Oct 2026Android gets AI defenses against messaging scams, fake calls
Google rolls out AI defenses on Android against scams, fake calls, and call forwarding abuse.
- 15 Oct 2026The next election influence campaign will target your AI chatbot
Influence campaign seeds internet with hyperlocal content to train AI chatbots favoring specific candidates.
- 15 Oct 2026Anthropic Warns Small AI Models Could Power Self-Spreading AI Worms
Anthropic warns small AI models could enable self-spreading AI worms targeting connected devices.
- 15 Oct 2026Why AI Confidence Is Not Validation
AI tools in forensics and incident response are outpacing human validation methods.
How we list vendors
- A product is listed because it provides the control. Nobody pays to be listed, and on story pages the examples rotate, so no vendor is always first.
- Vendor pages show both the vendor's own research and the incidents involving its products, from our own coverage. We do not leave either out.
- Our summaries are written by us, based on the vendors' own pages. Text in quotation marks or marked "in their own words" is the vendor's own, quoted from its website.
- Logos, product names and quoted text belong to the vendors and are shown to identify them.
- Something wrong? Use "Suggest a correction" on the vendor's page. A product we should look at? Email support@threatnoir.com.