Protecting The Core Of Generative AI
- Tweets 172
- Following 13
- Followers 21
- Likes 18
Pinned Tweet
Keep your favourite #CodingAgent. #Encrypt your #AIData.
It works with @claudeai Desktop/CLI, @OpenAI Codex, @opencode, @openclaw, @NousResearch Hermes Agent, @SarvamAI , and more.
If your agent speaks OpenAI's API, it already speaks #MirrorETF.
mirrorsecurity.io/etf
1
4
128
#AskCEO: Can your guardrails check a prompt they can't read?
Pankaj Thapa on encrypted guardrails: prompt injection, PII, bias and toxicity detection on encrypted input and output streams.
#AISecurity #AgenticAI #Guardrails #FHE #BlackHat2026 #MirrorSecurity
1
27
Does your guardrail see your prompt? Today, yes.
Every guardrail service reads the text to score it. Which means the safety check you added is one more place your data is exposed.
1
2
4
22
Answer: Encrypted guardrails. Runtime AI security on encrypted input and output.
#AISecurity #AgentSecurity #MirrorSecurity #encryptedGuardrails #guardrails
1
17
what is Fully Homomorphic encryption?
#FullyHomomorphicEncryption #AIDataExposureRisk #AISecurity #MirrorSecurity
ALT AI works on your data. It never sees it. What is fully homomorphic encryption? Mirror Security carousel cover.
ALT Goes in locked, comes out still locked. Sealed input, server computes blind, sealed result. No decryption point.
ALT The server adds blindfolded, your key reveals eight. Worked example: (15, 407) + (23, 664) = (38, 71); 71 minus 38 times 7 = 805 = 100 times 8 + 5.
ALT Compute on encrypted data. Decrypt only at home. Mirror ETF, mirrorsecurity.io.
2
4
17
MirrorSecurity retweeted
Google CyberSecurity Cohort @google @GoogleStartups London. Great sessions & Discussions !!!
2
1
6
105
#AskCEO: What is ETF?
Encrypted inferencing, encrypted memory, encrypted harness. Your AI agents work, your data stays protected. CEO Pankaj Thapa explains in 90 seconds.
mirrorsecurity.io/etf
#AISecurity #DataPrivacy #MirrorSecurity #FHE #EncryptedAI
2
26
Your prompt does not just disappear into the model. It sits in infrastructure, feeds training loops, and exposure goes beyond PII.
CEO Pankaj Thapa explains in under a minute: mirrorsecurity.io/etf
#AISecurity #DataPrivacy #MirrorSecurity #AIDataSecurity #AIDataExposureRisk
2
6
37
"What is Fully Homomorphic Encryption and how does it work?"
is the common question we get everywhere. Here you go!
#FullyHomomorphicEncryption #SoverignAI #PrivateAI
2
5
34
5 AI data security myths, busted at Black Hat 2026.
A pull-aside conversation with our cofounder Aditya Narayana, on the claims we hear at the booth every day.
Run inference on fully encrypted data → mirrorsecurity.io/etf
#BlackHat2026 #AISecurity
3
6
113
A few months ago, ours was just a plan, a 10×10 square and a drawing. Then it became a frame, a backdrop, a couple of bar stools… and then @BlackHatEvents opened the doors, and you showed up. Thank you, everyone, who visited us. #MirrorSecurity #AIDataSecurity #AISecurity
4
5
77
Mirror Security is hosting Securing Agentic AI: Your Data is the Target. Make sure to attend it on August 26. linkedin.com/events/securing…
1
3
27
Your prompts are your most sensitive data. Most of the time, you hand them over in plaintext. We wanted to fix that without asking anyone to leave the tools they already use. So we brought the Encrypted Token Factory inside Claude Desktop.
#EncryptedAI #ConfidentialComputing
1
1
4
53
Run encrypted inference on Claude Desktop with Mirror ETF. #AIsecurity #DataPrivacy #ClaudeDesktop #FHE #SovereignAI
→ mirrorsecurity.io/etf
1
14
One command switches the AI tools you already use from tokens to encrypted tokens. Your data stays unreadable, even during inference. No SDK. No migration. Founding seats are open.
Meet us at @BlackHatEvents
mirrorsecurity.io/etfearlyad…
#BH26 #AI #AIsecurity
3
5
108
Excited to be part of this! Over the coming months, we'll be building alongside experts from @Google Cloud, @GoogleDeepMind, Wiz & Chronicle, exploring the possibilities with the Gemini ecosystem and pushing what secure, privacy-preserving AI can actually do at enterprise scale.
Meet the 33 trailblazing startups joining the second Gemini Startup Forum: Cybersecurity. 🔒💻
This September, these global startup teams will take part in a unique two-day, in-person event in London that offers them a chance to engage in forward-thinking dialogue with AI experts from @Google.
Participants will collaborate directly with experts from across @GoogleDeepMind and @GoogleCloud to explore how AI will redefine the cybersecurity landscape. Founders will leave not just with insights, but with a tangible, actionable AI roadmap for their cybersecurity startup.
Learn more about the cohort in our blog: goo.gle/4vfbEc5
1
3
49
