What we are seeing across real AI estates: governance that survives an audit, attacks that actually land, and the operational control that closes the gap between the two.
DeepSeek Harness let a sandboxed agent grant itself full access with one curl command. Codex, Mistral Vibe and brig had their own bypasses. What went wrong, and where enforcement for coding agents has to live.
·6 min read
The FTC chair says an AI agent is a tool and the person instructing it is responsible. For teams running coding agents, that turns one question urgent: can you show who started a run, what it was allowed to do and what it did?
·7 min read
A user reports that an agent-written cleanup script deleted 48,218 live files and the local Git history after walking through 614 Windows junctions. What went wrong, and the controls that keep agent deletes inside the folder they were meant for.
·6 min read
DarkControl now tracks the token usage of your AI coding agents. Engineers see where tokens go and how to use fewer. Finance sees what agent usage adds up to. Both read it next to the audit trail they already trust.
·3 min read
Researchers found that most major coding-agent harnesses let agents delete their own traces without tripping monitors. Here is why agent audit trails have to be captured outside the agent, and what that means for liability and compliance.
·6 min read
OpenAI dots shipped allow, block and approval rules. NVIDIA announced an agent runtime boundary. Claude Code has managed permission rules. Each covers one agent. Here is why per-vendor rules are not an org policy, and what to put above them.
·6 min read
Coding agents at 300+ organizations pushed 13,000+ internal screenshots, including billing records and treasury consoles, to public GitHub repos so they would show up in code review. Nobody attacked them. Here is why it happened and which guardrails stop it.
·6 min read ·1 like
CVE-2026-90898 let one unauthenticated POST to the Bifrost AI gateway start any program on the host. The lesson goes beyond one product: a stdio MCP entry is a command line, and whatever can write MCP config can run code.
·6 min read
Coding agents run git status and git diff constantly. Manifold Security showed that a malicious core.fsmonitor setting turns those harmless commands into code execution across several agents. Here is what to change today, and how to keep visibility over what really runs.
·3 min read