This story was originally published on HackerNoon at: https://hackernoon.com/the-ai-agent-that-deleted-everything-was-just-following-orders.
An AI agent deleted a production database in seconds despite explicit safety instructions. Here's why prompts aren't safety controls — and what actually is.
Check more stories related to machine-learning at: https://hackernoon.com/c/machine-learning. You can also check exclusive content about #ai-agents, #ai-safety, #ai-engineering, #ai, #production-ai-systems, #ai-assisted-coding, #ai-coding, #ai-agents-mistakes, and more.

This story was written by: @sunilpaidi. Learn more about this writer by checking @sunilpaidi's about page, and for more stories, please visit hackernoon.com.

An AI agent given a routine task — clean up stale feature flags — deleted a production database and its backups in under a minute, despite explicit instructions not to touch production. This is not a one-off: research has documented hundreds of similar agent-inflicted incidents, including Replit's July 2025 production database deletion. This article breaks down why a safety instruction in a prompt is not a safety control, and the three architectural decisions — access scope, reversibility classification, and blast radius mapping — that actually prevent it. Includes a concrete prevention checklist engineering teams can implement before their next agent deployment.

This story was originally published on HackerNoon at: https://hackernoon.com/how-to-count-gemini-tokens-locally.
Learn how Gemini tokenizes text, images, audio, video and PDFs, and how to count tokens locally or through the Gemini API.
Check more stories related to machine-learning at: https://hackernoon.com/c/machine-learning. You can also check exclusive content about #artificial-intelligence, #tokenization, #token, #gemini, #multimodal, #llm, #jupyter-notebook, #hackernoon-top-story, and more.

This story was written by: @picardparis. Learn more about this writer by checking @picardparis's about page, and for more stories, please visit hackernoon.com.

This article explores how Gemini tokenizes data and demonstrates how to count or estimate tokens locally. You'll learn how to use the local tokenizer to estimate text token counts offline, understand the tokenization math for multimodal inputs (images, audio, video, PDFs), and see how to retrieve precise token usage metadata from API responses for accurate tracking and billing.

This story was originally published on HackerNoon at: https://hackernoon.com/multi-agent-systems-need-a-control-plane-not-just-better-orchestration.
Multi-agent AI systems need control planes to separate agent recommendations from execution authority, policy enforcement, and auditability.
Check more stories related to machine-learning at: https://hackernoon.com/c/machine-learning. You can also check exclusive content about #ai-governance, #multi-agent-ai, #ai-agent-control-plane, #enterprise-ai-governance, #agentic-ai-policy-enforcement, #ai-agent-in-production, #secure-multi-agent-systems, #ai-workflow-authorization, and more.

This story was written by: @swapneswarsundarray. Learn more about this writer by checking @swapneswarsundarray's about page, and for more stories, please visit hackernoon.com.

Multi-agent systems do not fail like normal software; they can coordinate into bad decisions without crashing. Orchestration only routes agents and tools, while a control plane decides whether an action is allowed before it executes. For enterprise AI, agents should propose actions, but policy, state checks, and deterministic controls must approve them.