Wildroot

Queries may use an external AI service. Details

Wildroot Browser

TodayStory on Mastodon

#AI

Posts
15
People
13
Time
–
Mostly
AI

canaries

All 15 posts of this story, best first: each person’s best post, then their other posts. The title is the name people mention most, or the opening line of the best post.

  1. "We sent nearly two million AI prompts containing unique, trackable secrets to AI providers, then watched where those secrets surfaced. What we found revealed security breaches, prompt theft, and widespread failures to protect customer data. Over a five-month period from July through November 2026, we submitted approximately two million prompts containing unique URLs and canary token credentials to widely used AI services and monitored what happened after inference. Our investigation uncovered three significant findings:

    Show more of this post

    - Evidence of provider compromises. - Unauthorized access to customer prompts by inference marketplace providers. - Failures in privacy and data-handling practices. We observed multiple" dragos.com/blog/ai-supply-chai… Privacy settings were ignored: "Zero Data Retention is supposed to mean something. In practice, these prompts were retained, shared with other companies, and read by people." What to do? - Host it yourself, if you can afford it. - Go directly to a first-tier provider. - Use routers with care (or not at all) - Seed your own canaries. - Enable secret filtering

    Info
  2. @9x0rg@mamot.fr

    The research is the latest indictment exposing the lax privacy and security that’s endemic in much of the AI ecosystem. The clearest example involves a Chinese AI lab. This lab’s systems fetched the secret URLs from more than half of the prompts that we sent to its models, including roughly two-thirds of the requests marked as “no data collection” or Zero Data Retention. **Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced.** arstechnica.com/security/2026/…

  3. @markhughes@hachyderm.io

    What happens in AI chat doesn't stay in AI chat. Oops, all your secrets are no longer secrets, a researcher found: - Evidence that at least three AI providers were compromised by attackers who used their access to siphon prompts they received - Multiple resellers in the inference marketplace that logged the prompts and somehow passed them along to people who used the credentials - Multiple cases of AI providers not following end-user privacy settings embedded into flags specifying terms such as Zero Data Retention (ZDR) or “no data collection” arstechnica.com/security/2026/… #LLM #Privacy #Security #AI

    Info
  4. In addition to the proof of the Riemann hypothesis we now also have the proof that AI service providers resell your conversations (all of it), despite claims otherwise. What a time to be alive. arstechnica.com/security/2026/…

  5. @miblo@mas.to

    🔒 Someone sent me a little single-issue security report for my personal website. The issue did turn out to be legit and I've resolved it, but the report itself bore the stench of LLM-generated slop. So, while I do appreciate being clued in about this, I'd rather not legitimise and encourage any LLM-use, and have decided not to reply with my thanks, thank you very much! ⚠️ The issue: A .htpasswd file accessible for anyone to download. ✅ The resolution: Keeping the .htpasswd file in a directory that isn't served to the web. And a bonus measure on this theme, denying access outright to certain files: location ~/\.(git|env|htaccess|htpasswd) { deny all; return 404; } #Security

  6. @evacide@hachyderm.io

    It is very difficult to be absolutely certain that your prompts are private, which is bad news for a lot of people who think they have privacy protections that they don't: arstechnica.com/security/2026/…

    Popular todayprobably AI
  7. Remember the Dodge idiots bringing in their personal laptops and hooking stuff up to secure networks? Remember the idiot using aClaude coding assistant while working on SIPRNet? We can safely assume every system is able to be compromised at this point because, idiots. arstechnica.com/security/2026/…

    Opinion
  8. @osma@mas.to

    Is it not incredibly obvious that the ethics compromised, hemorrhaging money, desperate for revenue purveyors of AI apps, routers, models and so forth will sell any and all user data they acquire to anyone bidding? Well, it's proven here. arstechnica.com/security/2026/…

  9. @jinxc@mas.to

    Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced. arstechnica.com/security/2026/…

    Info
  10. @HNguyenLy@mas.to

    Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced. - Ars Technica arstechnica.com/security/2026/…

    Info

More posts from the same people

  1. @osma@mas.to

    These are the same people who previously were creating data reselling ad networks, fishy app SDKs for "attribution", pushing crypto tokens, and outright scams. Of course they will take your prompts and sell them. You'd be a fool to have expected otherwise.

    Opinion