Urgent.News

What's breaking now, across thousands of outlets.

AI

Your AI agent just provisioned a resource. Who owns it?

The environment has thirty-four resources, including a VPC, subnets, an RDS instance, and a load balancer that hasn’t seen any The post Your AI agent just provisioned a resource. Who owns it? appeared first on The New Stack .

Your AI agent just provisioned a resource. Who owns it?

Your AI agent just spun up a resource. Determining who holds ownership of that resource poses a challenge. A departed employee leaves a trail behind, but agents leave none of that. Once their task concludes, their ownership vanishes, just like a candle extinguished. This issue compounds as agents operate in unpredictable ways, owning resources for fleeting periods.

The solution remains consistent: a query, a policy, and a guardrail. The query utilizes CloudQuery's asset inventory to identify the owner of a tagged resource. By cross-referencing this owner with your identity directory, any non-human account stands out. The policy prevents agents from owning themselves by checking the owner tag against your identity directory. If the owner isn't a human account, the policy flags it.

However, this process doesn't address resources already built and abandoned by agents. In such cases, assigning a last day to agents' resources ensures they comply with the ownership rules. The environment's TTL automatically eliminates resources that exceed this limit. This approach guarantees that even older resources adhere to the established ownership guidelines.

Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at thenewstack.io →

More in AI

Gemma 4 E2B on an AMD MI300X: Which Weight Format Should You Serve?

This article provides a step by step guide to serving ten weight formats of Gemma 4 E2B on one AMD Instinct MI300X through vLLM, with every build timed across a grid of request counts and prompt…

  • fp8 format matches bf16 speed on AMD MI300X, achieving 0.75x for single request
  • Int8 W8A8 format significantly slower, from 0.29x to 0.87x
  • fp8 format has largest deviation from Google's trained weights

Your Agent's Memory Is a Lottery: The Memory-vs-Documentation War of 2026

The hook: a page about everyone you know, refreshed hourly In the first week of October 2026, independent AI safety researcher Karan Joshi did something almost comically simple: he asked Meta's new AI…

  • Meta's Muse unveiled agent memory system in October 2026
  • Critics argue Muse creates dossiers without consent
  • Engineer Kevin Liao proposes documentation over memory

More from Thursday 8 October →