Every AI tool leaves invisible fingerprints in your text. SILK finds them — AI watermarks, hidden spaces, leaked credentials, personal data — and removes them before you share.
Every time you copy and paste, invisible data travels with your text. Most of it is harmless. Some of it isn't.
ChatGPT, Claude and Gemini embed zero-width characters — ZWS, ZWNJ, WJ — into generated text. Invisible to the eye, detectable by software. They flag your content as AI-generated even after editing and polishing.
Zero-width charactersBeyond invisible chars, AI tools create detectable patterns in word choice and token distribution. GPT-o3 and o4-mini embed narrow no-break spaces at 5× the human rate. Simple character removal won't fix this — our Pro tier uses Claude to neutralise token-level signatures.
Token-level patternsEmails, phone numbers, NI numbers, API keys, JWTs and card numbers travel silently in copied text. SILK finds them before they reach ChatGPT, Slack, or a client email, and replaces them with safe tokens that preserve document structure.
PII & secretsNon-breaking spaces, soft hyphens, smart quotes and em dashes copy invisibly between systems. They look normal in Word but break code, corrupt CSV exports, fail database inserts, and cause subtle rendering issues across CMS and email platforms.
Formatting artifactsEveryone has controls on email. Nobody has controls on paste.
You write a proposal with ChatGPT, paste it into a document and send it to a client. Unknown to you, zero-width characters mark every sentence. Their compliance team flags it as AI-generated. The deal is at risk.
Zero-width spaces are embedded by every major AI model during text generation. They survive copy-paste, editing and reformatting. They are invisible in Word, invisible in Google Docs, invisible in email — but detectable by automated compliance tools.
You copy a section from a board pack to paste into an AI tool. Three lines below your selection sits the CFO's salary and a live acquisition target. It goes to ChatGPT. It leaves your device. It's in a US data centre.
Under GDPR Art. 25 and FCA SYSC 6.1, you are responsible for data that leaves your environment — even accidentally. The clipboard has no audit trail. SILK creates one.
A developer copies a config snippet to share in Slack. Four lines below sits a live AWS access key. It reaches 40 people. The key is valid for three more months. Security ops has to rotate every affected service.
API keys, JWTs and access tokens are the highest-severity clipboard risk. A single paste to the wrong channel can expose your entire infrastructure. SILK detects and removes credentials before they leave the editor.
AI text, emails, CVs, contracts, code — anything you're about to copy somewhere else. SILK accepts all of it instantly.
Hidden characters appear as coloured markers inline inside your text at their exact position — you see the contamination before you fix it. Blue for AI watermarks. Amber for hidden spaces. Orange for personal data. Red for credentials.
One click removes everything. PII is replaced with safe tokens. The editor updates in place. Download a compliance PDF for your audit trail. Copy the clean version. Revert any time.
All the essentials for daily clipboard hygiene. No account required.
AI-powered features for professionals handling sensitive content daily.
White-label, on-premise, dedicated compliance documentation and support.