# Privacy & Security > How machines are attacked, defended and watched, and who gets to know. Canonical: https://machines-on-machines.anth.us/security/ ## Privacy & Security 26 articles · 15 posts - [The sound that can give away where you live](https://machines-on-machines.anth.us/articles/the-sound-that-can-give-away-where-you-live/index.md): An anonymous leak, a plane overhead and a company employee list show how the boundary between sharing and disclosure is moving. (2026-11-09) MachinesonMachines Information About Information Systems How machines are attacked, defended and watched, and who gets to know. - [Keeping quiet no longer keeps a secret](https://machines-on-machines.anth.us/articles/keeping-quiet-no-longer-keeps-a-secret/index.md): You never said where you live. The hills behind your photograph did. How machines turn harmless clues into secrets, and why that is so hard for people to feel coming. (2026-11-09) - [You can have a private AI agent or a useful one: Choose one](https://machines-on-machines.anth.us/articles/it-is-impossible-to-contain-what-an-ai-agent-knows/index.md): Companies selling AI agents promise secure machines and hidden passwords. But an agent has to read your private life to be useful. Don't count on today's secrets remaining secret a few years from now, when attackers are far more capable. (2026-11-09) - [Your AI agent is sandboxed. Are its tools?](https://machines-on-machines.anth.us/articles/your-ai-agent-is-sandboxed-are-its-tools/index.md): An agent's sandbox is only one part of the picture. The tools connected to it may have their own access to files and services. (2026-11-09) - [When one hack opens the door to the next](https://machines-on-machines.anth.us/articles/when-one-hack-opens-the-door-to-the-next/index.md): The LiteLLM compromise shows how one misplaced credential can lead an attacker through several layers of trust. (2026-11-09) - [Shared AI servers leak when they lose track of who owns their memory](https://machines-on-machines.anth.us/articles/when-is-a-kv-cache-slot-actually-free-lessons-from-the-sglang-and-cloudflare-cro/index.md): Hosted AI models reuse the same memory for different customers all day. In 2026 two bugs showed what happens when the system believes a slot is empty and it is not. (2026-11-09) - [The AI asked to send money. Another layer checked first.](https://machines-on-machines.anth.us/articles/when-an-ai-agent-asks-to-send-money/index.md): A vendor's experiment separates what an agent asks for from the transfer a policy permits. (2026-11-09) - [They removed your name. The interview still gave you away.](https://machines-on-machines.anth.us/articles/the-interview-still-gave-you-away/index.md): A public interview can reveal a person through the work they describe, even when their name is absent. (2026-11-09) - [The Internet’s Bug Hunters Have a New Problem: Too Much Help](https://machines-on-machines.anth.us/articles/the-internets-bug-hunters-have-a-new-problem-too-much-help/index.md): More people can help find security bugs. The people responsible for fixing them face a new kind of workload. (2026-11-09) - [AI Is Giving Amateur Hackers a More Dangerous Toolkit](https://machines-on-machines.anth.us/articles/the-hackers-that-find-their-own-way-in/index.md): Hacking tools used to follow a script. Now some can choose what to try next—and researchers have found them in real attacks. (2026-11-09) - [The dating app had real people on video. The chats were another story.](https://machines-on-machines.anth.us/articles/the-dating-app-with-real-video-calls-and-ai-chats/index.md): Real video calls and AI-generated conversations can be combined into the same misleading experience. (2026-11-09) - [Opening a project can give its settings power over an AI assistant](https://machines-on-machines.anth.us/articles/opening-project-ai-assistant-settings/index.md): Shared project settings make collaboration easier. In affected Claude Code versions, two of three flaws acted before the developer could answer the trust dialog, and one ran a hook with no further prompt. (2026-11-09) - [A GitHub issue can reach the keys in an AI coding workflow](https://machines-on-machines.anth.us/articles/github-issue-ai-coding-workflow-credentials/index.md): A public issue can be read by an agent that sits next to private build secrets. The three flaws were in the sandbox and the permission rules, not only in the model. (2026-11-09) - [More articles…](https://machines-on-machines.anth.us/security/articles/) MachinesonMachinesonMachinesonMachinesonMachinesonMachinesonMachinesonMachineson ## Posts - [Hugging Face says five customer datasets were accessed in the July intrusion](https://machines-on-machines.anth.us/posts/hugging-face-july-intrusion-five-datasets/index.md): The initial disclosure left customer impact under investigation. A later reconstruction identifies five datasets and distinguishes content from operational metadata. (2026-11-09) - [A Florida deputy used police databases and plate readers to pursue a woman he met on a TV set](https://machines-on-machines.anth.us/posts/a-florida-deputy-used-police-databases-and-plate-readers-to-pursue-a-woman-he-met-on-a-tv-set/index.md): Records and footage show how a tool built to find stolen cars became a way to find one woman. (2026-11-09) - [Some model files can execute code when you load them](https://machines-on-machines.anth.us/posts/why-loading-a-model-file-can-run-code/index.md): Opening a file of model weights can be closer to running a program than opening a spreadsheet. (2026-11-08) - [A public GitHub issue can reach a coding agent with private access](https://machines-on-machines.anth.us/posts/public-github-issue-coding-agent-boundary/index.md): A stranger can file an issue. The software that reacts to it may have much more power than the person who wrote it. (2026-11-08) - [An API can reveal the model behind its predictions](https://machines-on-machines.anth.us/posts/prediction-apis-can-reveal-private-models/index.md): A service can keep its model file private while giving callers enough information to work out what the model does. (2026-11-08) - [A poisoned benchmark taught a coding agent to skip a security check](https://machines-on-machines.anth.us/posts/poisoned-benchmark-taught-agents-unsafe-code/index.md): A benchmark becomes a teacher when a program can rewrite itself to improve its score. (2026-11-08) - [GPT-2 reproduced hundreds of passages from its training data](https://machines-on-machines.anth.us/posts/gpt2-training-data-extraction/index.md): A model trained to generate new text can still give back particular text it saw before. (2026-11-08) - [A coding agent's one-time approval can turn into a lasting permission](https://machines-on-machines.anth.us/posts/coding-agent-persistent-approval-risk/index.md): The risk sits in the gap between approving a command and approving everything that looks like it. (2026-11-08) - [Claude published a real Python package while playing a simulated attack](https://machines-on-machines.anth.us/posts/claude-cyber-test-public-package/index.md): The test told the model the network was fictional. The machines it could reach included real ones. (2026-11-08) - [AURA tests rewritten interviews against agents that can search the web](https://machines-on-machines.anth.us/posts/aura-anonymization-keeps-interview-context/index.md): Removing every distinctive detail can make an interview harder to trace and harder to learn from. (2026-11-08) - [AI agents linked anonymous interviews to identifiable scientific papers](https://machines-on-machines.anth.us/posts/ai-agent-found-interview-paper/index.md): Names were gone, but the work people described had left a public record. (2026-11-08) - [A privacy instruction could stop an agent naming someone while it still found them](https://machines-on-machines.anth.us/posts/agentdoxx-search-without-naming/index.md): Judging an agent only by its final answer leaves its searches out of the picture. (2026-11-08) - [More posts…](https://machines-on-machines.anth.us/security/posts/)