Handling PII in customer-facing AI chatbots: mask before sending to LLM

Comments (1)

hoverbot · 3h ago

Author here (HoverBot team). How we keep PII out of LLMs in our chatbots: detect fast -> replace with typed placeholders -> call the model -> unmask on our server. We also cover region pinning and vendor privacy controls. Keen to hear trade-offs, failure modes, and better designs.

AI 'Map Reduce': Scaling AI Tasks (danielsada.tech)

The Centenary of the Loudspeaker (nickhunn.com)

Show HN: RemoteMCPList – A community directory for MCP servers

Junior devs not interested in software engineering (old.reddit.com)

GitHub hides "cyberpunk" and "cyberpunk2077" topics (github.com)

New CRISPR technique could rewrite future of genetic disease treatment (unsw.edu.au)

The Illumos Cafe: Another Cozy Corner for OS Diversity (it-notes.dragas.net)

Chat Control is back and we've got two months to stop the EU CSAM scanning plans (tuta.com)

Miniforge: Community-led conda-forge distribution (github.com)

Ofcom £20k fine of Delaware-incorporated 4chan is illegal says US law firm (mobilenewscwp.co.uk)

The Nvidia AI GPU black market [video] (youtube.com)

A Trick to Use MkMerge at the Top Level of a NixOS Module (gist.github.com)

Spritacular – a citizen science project of Transient Luminous Events (TLEs) (spritacular.org)

Parquet Content-Defined Chunking (huggingface.co)

Show HN: We made a local-first LLM assisted journaling tool (github.com)

The joy of recursion, immutable data, & pure functions: Making mazes with JS (jrsinclair.com)

Taiwan's "silicon shield" could be weakening (technologyreview.com)

BriskBard 4.6 Released (briskbard.com)

Ask HN: What's your go‑to for mapping cross‑team dependencies?

Show HN: I used LLMs to build an OS AI voice agent that guarantees accurate data (github.com)

Mindless Machines, Mindless Myths (lareviewofbooks.org)

Show HN: Founderly – an AI cofounder to take you from idea to launch

How Will AI-Driven Learning Platforms Reshape Enterprise Upskilling?

Meta's DINOv3: Self-supervised learning for vision at unprecedented scale (ai.meta.com)

The Cost of Slow Feedback Loops (revontulet.dev)

I talked to Sam Altman about the GPT-5 launch fiasco (theverge.com)

Software Decoding and the Future of Mobile Video (streamingmedia.com)

Philips Hue's new bridge could turn your lights into motion sensors (theverge.com)

AI That Customizes Every Email-Is This Worth Building?

Find MCP Servers. Build AI Agents Quickly (mcp.so)

AlphaEarth Foundations: a universal embedding for Earth observation data (newsletter.caffeinatedengineer.dev)

AI That Customizes Every Email-Is This Worth Building?

What's the Difference in Token Formats–and Why It Matters (ixopay.com)

How to Make the Most of Veo3, WAN 2.2, Hailuo-AI, and Qwen-Image on Textideo (indiehackers.com)

Unification (2018) (eli.thegreenplace.net)

Always Winning: Reverse engineering a festival app's mini-games with Frida (kopanko.com)

'Tradwife', 'delulu' and 'skibidi' among new words added to Cambridge Dictionary (news.sky.com)

A short statistical reasoning test (emiruz.com)

Model Evaluation (ampcode.com)

Ask HN: How do you set Newsletter pricing? Confused 8k

From Monolith to Cloud: Automating Your Migration Journey (blog.qaware.de)

Finding a Successor to the FHS (lwn.net)

The last time you experienced pure, undistracted music – how long ago? (serenesound.app)

Climate change makes South Asia's monsoon season more dangerous (apnews.com)

Show HN: A tool to discover rapidly rising AI open source projects (trickle.so)

Sam Altman says 'yes,' AI is in a bubble (theverge.com)

When Philosophy Meets AI (github.com)

Web apps in a single, portable, self-updating, vanilla HTML file (hyperclay.com)

EloqKV, a distributed database with Redis compatible API (GPLv2 and AGPLv3) (github.com)

Wan 2.2, a new Video generation AI model (wan.video)

Handling PII in customer-facing AI chatbots: mask before sending to LLM

Comments (1)