> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Headline:

69% of Enterprise AI Agents Share Login Credentials Like Teenagers
- URL: https://wire.fourthweb.ai/headline-69-of-enterprise-ai-agents-share-login-credentials-like-teenagers/
- Published: 2026-07-29T20:02:03.000Z
- Updated: 2026-07-31T04:02:45.000Z
- Description: The agent economy has a plumbing problem — and it's not the kind you can fix with better prompts. 69% of enterprises run AI agents that share credentials, a practice linked to higher security incidents, according to VentureBeat research
- Author: Travis Wright
- Tags: AI Agent Economy, Agentic Workflows, AI Agents, Institutional Crypto, Anthropic, IPO Watch

**The agent economy has a plumbing problem — and it's not the kind you can fix with better prompts.**

### The Summary

- [69% of enterprises run AI agents that share credentials](https://venturebeat.com/security/ntt-data-aivista-and-snowflake-identity-alone-wont-secure-enterprise-ai-agents?ref=wire.fourthweb.ai), a practice linked to higher security incidents, according to VentureBeat research
- [Agents can't see each other, can't coordinate tasks, and leave no tamper-proof audit trail](https://venturebeat.com/orchestration/enterprise-ai-agents-cant-talk-to-each-other-cant-be-trusted-with-permissions-and-cant-be-audited-5-startups-are-already-fixing-that?ref=wire.fourthweb.ai) when they act autonomously
- Infrastructure startups like BAND are building coordination layers so agents can recruit peers, delegate subtasks, and return results without human handholding
- The real problem isn't identity — it's action-level authorization and proving to auditors that agents did what they said they did

### The Signal

Enterprises are deploying agents that can book flights, approve expenses, and query databases. But those agents are operating in what amounts to digital solitary confinement. They can't talk to each other. They share login credentials like roommates sharing a Netflix password. And when something breaks, [there's no tamper-resistant record of what actually happened](https://venturebeat.com/security/ntt-data-aivista-and-snowflake-identity-alone-wont-secure-enterprise-ai-agents?ref=wire.fourthweb.ai).

At VB Transform 2026, executives from NTT DATA AIVista and Snowflake laid out the problem. Mukesh Karki, CTO of AIVista, put it bluntly: "These organizations need to be able to prove to their auditors in a very tamper-resistant fashion that those records showing what they did actually prove what they're doing. And the provability is essentially your license to operate in a regulatory environment."

> "In the agentic world, the software has a brain of its own, and it's constantly rewiring itself."

The issue is rooted in assumptions from an earlier era of software. [Mayank Upadhyay, Snowflake's chief security and trust officer, explained](https://venturebeat.com/security/ntt-data-aivista-and-snowflake-identity-alone-wont-secure-enterprise-ai-agents?ref=wire.fourthweb.ai): "In the traditional software world, a human being clicks somewhere and the software does something very deterministic, and you know which API it's going to call. But in the agentic world, the software has a brain of its own, and it's constantly rewiring itself." Give an agent too much permission and it will explore. Agents are curious. They try things. And [when they share credentials, unintended side effects multiply](https://venturebeat.com/security/ntt-data-aivista-and-snowflake-identity-alone-wont-secure-enterprise-ai-agents?ref=wire.fourthweb.ai).

Identity alone won't fix this. You need action-level authorization, meaning the agent can only do what it's supposed to do right now, not everything the human whose credentials it borrowed could theoretically do. You need audit trails that can survive scrutiny from regulators who assume you're lying. And you need a way for agents to actually coordinate.

That's where the infrastructure layer comes in. [BAND, one of five startups showcased at the conference](https://venturebeat.com/orchestration/enterprise-ai-agents-cant-talk-to-each-other-cant-be-trusted-with-permissions-and-cant-be-audited-5-startups-are-already-fixing-that?ref=wire.fourthweb.ai), is building what Vlad Luzin, the company's CTO, calls a coordination infrastructure layer. The pitch: agents should be able to receive tasks, visit registries of other agents, recruit help, delegate subtasks in a conversational space, gather results, and return a summary to the human who kicked things off.

Why can't Slack or Discord handle this? [Because those platforms were built for humans](https://venturebeat.com/orchestration/enterprise-ai-agents-cant-talk-to-each-other-cant-be-trusted-with-permissions-and-cant-be-audited-5-startups-are-already-fixing-that?ref=wire.fourthweb.ai). Agents have to be onboarded manually. They can't see each other. [Claude](https://wire.fourthweb.ai/tag/anthropic/) is stateless. Developers toggle between multiple sessions for different tasks. Luzin frames it as a distributed systems problem: "The transportation layer needs to be solved first, how the agents communicate."

Key infrastructure gaps being addressed:

- Orchestration: How agents find, recruit, and coordinate with each other
- Observability: Tamper-proof logs of what agents actually did
- Authorization: Action-level permissions that expire when the task ends
- Connectivity: A transport layer that lets remote processes talk without human wiring

### The Implication

If you're running agents in production today, you're likely doing it with duct tape and shared API keys. That works until it doesn't. The smart move is to start thinking about infrastructure now, before your CFO's agent accidentally approves a six-figure invoice because it got curious. Watch for companies building coordination layers, authorization frameworks, and audit-ready observability tools. These are the picks and shovels of the agent economy.

The real test will be regulatory. When an agent screws up in a regulated industry, can you prove to an auditor what it did and why? If the answer is no, you don't have an agent strategy. You have a liability.

### Sources

[VentureBeat](https://venturebeat.com/security/ntt-data-aivista-and-snowflake-identity-alone-wont-secure-enterprise-ai-agents?ref=wire.fourthweb.ai) | [VentureBeat](https://venturebeat.com/orchestration/enterprise-ai-agents-cant-talk-to-each-other-cant-be-trusted-with-permissions-and-cant-be-audited-5-startups-are-already-fixing-that?ref=wire.fourthweb.ai)