Company data in AI tools
A customer email, an API key or internal project details end up in a prompt fast. Cortex Guard detects them and warns, redacts or blocks.
Cortex Guard catches hidden instructions, personal data and exposed credentials before they reach ChatGPT, Claude or Gemini. And enforces your rules.
Chrome & Edge · ChatGPT, Claude, Gemini
HAPPYBF2026
Applied automatically at checkout. If it isn’t, enter this code to get 30% off.
Ends December 3, 23:59 Berlin time
Secure payment via Stripe. Set up in 5 minutes. Chrome and Edge. Licenses bought in this offer are non-refundable.
Write a friendly payment reminder to Mr. Becker. His account number:
Your policy warns on bank details. Sending is allowed with a reason.
Review this supply agreement for unfavorable clauses.
Possible prompt injection. Nothing was sent to ChatGPT.
Skill tries to leak credentials. Upload to Claude stopped.
The Grammarly for AI security: Cortex Guard checks what your team sends to AI before it gets there.
A customer email, an API key or internal project details end up in a prompt fast. Cortex Guard detects them and warns, redacts or blocks.
A supplier PDF can contain white text that steers the AI. People don’t see it. The AI reads it.
Skills can try to leak confidential data or hide actions. Cortex Guard checks them before upload.
of the agent skills studied had vulnerabilities.
of skills showed likely malicious intent.
higher average cost of a data breach when shadow AI use is high.
Sources: NVIDIA SkillSpector study (June 2026), IBM Cost of a Data Breach.
The Black Friday offer ends December 3, 2026, at 11:59 pm Berlin time. Business customers only, non-refundable.
Buy directly, start right away
per seat per year, plus applicable tax. Renews at the regular price of $120, cancel anytime before renewal.
HAPPYBF2026
Applied automatically at checkout. If it isn’t, enter this code to get 30% off.
Central rollout with us
Deployment via Chrome Enterprise, custom rules and confidential topics, admin view for approvals. The Black Friday price applies here too.
Pick a file and send it to the AI. Once without Cortex Guard, once with it.
Nothing sent yet. Turn Cortex Guard off or on, then click “Send to AI”.
Simulated demo with sample files. Nothing is sent to a real AI. Whether an AI follows a hidden instruction is unpredictable: sometimes it does, sometimes it doesn’t. Cortex Guard checks first.
Yes, in their browser versions in Chrome and Edge, plus uploads, knowledge bases and skills in TextCortex. Desktop apps and Firefox are not supported.
Hidden instructions (prompt injection), personal data such as email addresses or bank details, credentials, and your own confidential topics, by keyword and by meaning.
No. No scanner catches every prompt injection. Cortex Guard is one layer of protection alongside training and clear policies. It is not legal advice or a privacy review.
No. For an approval request, the admin sees the file name, the findings and the person’s reason, not their chats.
Yes. In monitor mode, Cortex Guard reports findings but blocks nothing. You turn on enforcement per platform.
Your license renews at the regular price of $120 per seat per year. You can cancel anytime before the renewal date.
Yes, the invoice is emailed right after purchase. You can add your tax ID at checkout.
No. Licenses bought in the Black Friday offer are non-refundable. The offer is for business customers only. Try the demo above before you buy.
$84 instead of $120 per seat per year. Through December 3, 2026 only.