On Oct. 7, 2026 DuckDuckGo added encrypted link sharing to Duck.ai — a small feature with outsized practical value for marketing teams that already use chat-based AI. It doesn’t make models safer or smarter, but it reduces a common operational risk: unintentionally exposing campaign data, customer details or unreleased strategy through a shared chat link.
What the feature does
Duck.ai’s sharing creates a link that lets someone open a conversation — or a single AI reply — and continue it without signing in. The key privacy guarantee is client-side encryption: the conversation is encrypted on the user’s device before it’s uploaded to DuckDuckGo’s servers. The generated URL contains an identifier and a decryption key; the key sits in the URL fragment (the part after the “#”), which browsers typically don’t transmit to servers. That means DuckDuckGo can host and serve the encrypted data without holding the ability to decrypt it.
Recipients who continue a shared chat create a separate thread; new messages are not merged into the original conversation. DuckDuckGo also applies practical limits: links are long and randomized, shared pages include instructions not to be indexed by search engines, and links expire by default after 48 hours (users can choose one, seven or 30 days). Users can revoke access manually before expiry. Uploaded files are excluded from shared conversations, though generated images created inside a chat may be included. DuckDuckGo says it removes the encrypted conversation from its servers once a link expires.
What it protects — and what it doesn’t
This design addresses a frequent operational leak: chats intended for internal review becoming discoverable. In 2025, publicly shared chatbot conversations from other providers surfaced in search results, demonstrating how sharing defaults and discoverability settings can accidentally publish sensitive content. By keeping the decryption key client-side and limiting link lifetime, Duck.ai reduces the window and likelihood of unintended disclosure.
But the announcement stops short of cryptographic details and independent verification. DuckDuckGo describes the client-side encryption workflow and the retention policy, but it does not publish a technical audit or the specific algorithms and implementation details. For organizations with strict compliance or contractual requirements, the feature should be treated as a risk-mitigating control rather than a certified guarantee.
Three practical steps marketing teams should take
Duck.ai’s change is useful, but it works best inside governance and process. At minimum, marketing teams should:
1) Make privacy a formal evaluation criterion. When selecting AI tools, document where input data is stored, whether it can be used for model training, who can access shared content, and what sharing controls exist. Ask vendors whether client-side encryption is used, and whether sharing features exclude uploaded files or attachments.
2) Add simple sharing rules to workflows. Encrypted links lower exposure, but they don’t replace data hygiene. Remove unnecessary customer identifiers, avoid pasting confidential product roadmaps or unpublished financials into chats, review content before sharing, and limit recipients to those with a clear need. Treat shared links like a temporary access pass and log when sensitive conversations are distributed.
3) Use privacy features as procurement signals. For agencies and vendors, concrete privacy controls — short-lived encrypted sharing, explicit retention policies, and revocation options — matter to buyers. When privacy is material to adoption, request written explanations of how sharing works and include them in vendor assessments or contract language.
What to watch next
Product designers and security teams will be watching for independent audits or published technical disclosures that validate the client-side encryption claims. Enterprises should also evaluate how short-lived encrypted links fit with existing data-loss prevention (DLP), archiving and e-discovery requirements. Finally, expect procurement conversations to include not just privacy statements but feature-level evidence: demonstrable controls matter more than promises.
Duck.ai’s encrypted sharing isn’t a radical AI capability, but it is a practical example of building privacy into collaboration. For marketing teams, the immediate value is operational: fewer accidental exposures and clearer controls when sharing AI-generated work. The longer-term value will depend on the transparency and verifiability of the implementation, and on how organizations fold short-lived encrypted sharing into governance and vendor evaluation.
