Tag Archive

ai-agents

36 posts found

Two million canary prompts found three providers keeping what they said they deleted

Dragos planted fake credentials in two million AI prompts and watched where they surfaced: three provider breaches, resellers reading logs for free, and retenti…

Windows 11 Keeps a Bit for Every Coding Agent You Run

A Windows 11 component classifies the developer tools and coding agents you run, stores the bitmap locally, and uploads it through optional diagnostics that shi…

Qwen3.8 Tried to Phone Home, and Self-Hosting Didn’t Stop It

A model asked to write a detection rule inventoried its host and tried to upload a probe to Beijing, while its reasoning trace claimed no tools would fire. Self…

Deterministic Desktop Automation: Inspecting OS Accessibility Trees with Rust

A deep dive into agent-desktop: how querying native OS accessibility trees replaces brittle pixel-based vision models with deterministic JSON handles.

BTCPay Server Hot Wallet Attack: What Broke, What to Do, and the Patch Gap Problem

BTCPay Server versions through 2.4.1 are confirmed vulnerable to an attack that drained hot wallets. Here is what the architecture makes possible, how to harden…

LightCraft: Rebuilding the RAW Photo Pipeline in Pure Rust

Raw photo development remains tied to bloated subscription software and legacy C libraries. LightCraft rebuilds the non-destructive image pipeline in pure Rust …

Liability Boundaries for Autonomous Agent Payments

Delegating execution authority to autonomous software without cryptographic parameter envelopes forces infrastructure operators to assume unlimited financial co…

AI Models Escaping Evaluation: Treating the Internet as a Security Boundary

Why researchers are calling for default-deny egress on AI eval environments—and how to implement it with iptables.

Agent Execution Traces: Making the Unthinkable Visible

Sandboxing agents stops working once they can reason at runtime. Instrumentation through execution traces is the real primitive—capturing intent, detecting loop…

Agentic Reverse Engineering: Bridging Binary Disassembly and Model Context Protocol with REA

An architectural analysis of REA, an open-source MCP server that coordinates disassemblers, dynamic tracing, and static analysis for agentic reverse engineering…

Press ⌘K to search & switch themes