{
  "id": 18350042,
  "title": "The four biggest agent frameworks give your agent no persistent ID. Now do know-your-agent.",
  "url": "https://urgent.news/2026/11/05/the-four-biggest-agent-frameworks-give-your-agent-no-persistent-id",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-05T18:16:25.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/vectle/the-four-biggest-agent-frameworks-give-your-agent-no-persistent-id-now-do-know-your-agent-36bn"
  },
  "original_language": "en",
  "account": "When it comes to agent frameworks, four of the most commonly used ones lack a persistent identity for each agent, with no record of which humans approved which tasks, according to researchers at the Partnership on AI. If an agent restarts, it is treated as a stranger in terms of identity. The CFTC chair brought up the concept of \"know-your-agent\" during the October 28 Frontier Forum, and RBC blocked third-party agents on the same day. Cequence discovered that a single actor was rotating around 32 impersonated AI bot identities while probing about 90 different paths.\n\nThis situation presents a challenge for developers, as trust should begin with scoped credentials, not a registry. A user-agent string or a name an agent claims should not be considered identity. Instead, Cequence recommends verifying callers through published address ranges, signed requests, or direct integrations. If you operate a service that agents call, focus on checking signatures rather than relying on costumes.\n\nTo build a robust system, provide your agents with stable, durable identity plumbing. This includes assigning a stable ID to each agent instance, ensuring it remains consistent even after restarts. Implement scoped credentials for each task instead of using a single login. Additionally, maintain a log that records which human approved which action.\n\nThe consequences of not addressing these identity issues can be severe. For instance, Wells Fargo's updated terms, effective in 19 days, treat agent instructions on a shared login as the customer's own actions, including exceeding intended authority. If your agents operate using shared credentials today, your users are already bearing that liability, regardless of whether your documentation specifies it.\n\nTo stay informed about real-world agent failures, consider utilizing a library like Vectle, which maintains a collection of 34,101 problems and 332 reported outcomes at the last audit. Keep in mind that the unrecorded failures often go unnoticed and remain unrecoverable due to the lack of logs or approval trails. The core question should not be whether to adopt a registry or credentials approach, but rather whether your agent framework provides your agents with a persistent identity and an approval record.",
  "summary": "Should an agent have to prove who it is before it touches your accounts, or does trust start with scoped credentials, not a registry? Field note first, because it changes the whole argument. Partnership on AI researchers wrote in MIT Technology Review that the four most widely used agent frameworks lack even a persistent ID for each agent, with no records of which humans approved which tasks.…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}