Blog
Thoughts on encryption, privacy, and ciphertext that doesn't betray you when it leaks.
-
19 June 2026
What the decoy realism engine generates
Shape-correct, validator-passing decoys across 69 credential types: API keys, crypto wallets, healthcare records, bank and securities identifiers, government IDs. What the engine produces, and the checksums it has to pass.
Realism Engine Decoys -
29 May 2026
Detect AI-agent prompt injection in 5 seconds, at 1M-row scale
Bulk decoy tripwires for agents that read attacker-controlled text. Arm one tripwire per customer row with one CLI command, up to 1M rows, and page Slack, PagerDuty, or Datadog when one fires.
AI Agents Tripwires -
19 May 2026
Our compliance roadmap: certified today, audited post-launch
Cyber Essentials Certified today. SOC 2 Type II and ISO 27001 controls operating but not yet audited. Independent crypto audit and pentest on the post-launch funding window. The honest distinction between certified, audited, and operating, and the timeline we are running.
Compliance -
12 May 2026
Mini Shai-Hulud and deny.sh
On 11 May 2026 an attacker published 84 malicious versions across 42 @tanstack/* npm packages. deny.sh is not affected: no compromised package in our lockfiles, no
Security Supply chainpull_request_targetin our workflows, every publish done by hand from one machine. Here is the audit and what we are doing as a result. -
13 May 2026
The decoy realism engine
A static template pool is a fingerprint. We replaced ours with a 69-type generation engine that produces shape-locked, type-aware, AI-validated decoys per request, with a deterministic validator that rejects any candidate that would have authenticated as real. Decoys you couldn't write yourself.
Engineering Infrastructure -
12 May 2026
Why deniability needs infrastructure, not just a library
A 200-line SDK can give you deniable encryption. It cannot give you deniability. The difference is the operational layer around it: per-tenant isolation, audit posture, key custody, disclosure pipeline. That layer is infrastructure, and it is what deny.sh actually ships.
Category Positioning -
11 May 2026
How to verify deny.sh
A deniability product gets held to a higher bar. Four layers you can verify yourself in under fifteen minutes: the threat model, the cryptographic construction, the disclosure pipeline, and the supply chain. With concrete commands at each step.
Trust Verification -
Scheduled · 4 July 2026
Why we built deny.sh, the deniability infrastructure
The launch post, queued for the open-beta morning. Three pillars (Encrypt, Operate, Verify), four trust anchors, agents-first from day one. What deniability infrastructure means, who it's for, and how to verify any of this yourself.
Scheduled Open Beta Deniable Encryption -
Scheduled · Open beta week 2026
We run deny.sh on deny.sh
A credible security vendor uses its own product, on its own data, in production. We migrated a real operator credential under the deniable primitive we ship and ran three independent attack drills against it. Wrong-password observability, Argon2id cost on production hardware, and at-rest indistinguishability between real and decoy control files.
Scheduled Dogfooding Attack Drills -
April 2026
Your AI agent handles secrets. Does it handle them well?
AI agents are handling API keys, credentials, and sensitive data. Plaintext logs, context windows, and persistent memory make those secrets leak-prone. MCP plus deniable encryption gives agents a safer storage path.
AI Agents Enterprise -
April 2026
The $5 wrench problem and the leaks that surround it
Cryptography stops working the moment a motivated adversary can keep asking you for the key. Deniability is what lets you hand one over and still keep the secret. The classic xkcd wrench, answered.
Privacy Threat Model -
April 2026
Why deniable encryption matters more than you think
Standard encryption protects bytes from people who don't have the key. It does not stop leaked ciphertext from announcing that real secret material is present. That's the gap deniable encryption fills.
Privacy Encryption -
April 2026
Whitepaper: Deniable Encryption via XOR-Composed Control Data
The full technical specification. Algorithm, security analysis, threat model, empirical verification results, and honest discussion of limitations.
Whitepaper Research