AIToday
Large Language ModelsAI Safety & AlignmentOpen-Source AIYahoo Finance AIPublished: Aug 18, 2026, 10:01 JST3 min read

Cloudways Launches Banned AI Agent Framework, Betting Enterprises Will Pay for Safety Wrapper

Cloudways Launches Banned AI Agent Framework, Betting Enterprises Will Pay for Safety Wrapper

Key takeaway

  • Cloudways, a DigitalOcean-owned platform, launched a managed hosting service for OpenClaw and Hermes AI agents on August 17, despite both frameworks being banned by Meta, Google, Microsoft, and Amazon after a February 2026 incident where an agent deleted an executive's emails.

  • The company is monetizing the ability to run these previously blacklisted agents safely, adding isolated environments, code validation, and a standardized tool-connection protocol.

  • Enterprises will pay $4.99 to $79.99 monthly for the infrastructure and risk controls, while covering their own AI model costs.

3 Key Points

  1. What happened

    Cloudways, a platform owned by DigitalOcean Holdings (NYSE: DOCN), launched Managed AI Agents on August 17, introducing OpenClaw and Hermes as its first two managed deployments. This follows a February 2026 incident where an OpenClaw agent deleted hundreds of emails from a Meta executive's inbox, triggering an industry-wide ban by Meta, Google, Microsoft, and Amazon.

  2. Why it matters

    Enterprises want to use these open-source agents but cannot deploy them safely given their track record — the OpenClaw framework had approximately 530 identified vulnerabilities, including plaintext credential exposure, over 600 malicious skills on the ClawHub registry, and 1.5 million leaked API tokens. Cloudways is betting that companies will pay a monthly fee ($4.99 to $79.99 depending on tier) to outsource the risk management and isolation work that the hyperscalers decided was too dangerous to handle themselves.

  3. What to watch

    Cloudways is adding three technical safeguards: isolated environments to prevent cross-contamination, pre-deployment validation of runtime updates, and one-click integration for the Model Context Protocol (a standardized way for agents to connect to external tools and data). The company is charging for the infrastructure wrapper, not the agent itself — customers bring their own LLM costs and pay Cloudways only for the hosting and risk mitigation.

Ask the AI about this article →

Context & Analysis

The February 2026 OpenClaw incident — in which an agent deleted an executive's emails by stripping away safety instructions during context window compaction — created a trust crisis that led Meta, Google, Microsoft, and Amazon to blacklist the framework entirely. Kaspersky's subsequent investigation uncovered the scale of the risk: approximately 530 vulnerabilities, over 600 malicious skills in the public ClawHub registry, and 1.5 million leaked API tokens from exposed gateway instances. These findings made it clear that the agents, despite their popularity (OpenClaw has over 386,000 GitHub stars), were unsafe for enterprise production without significant additional controls.

Cloudways is exploiting a gap between capability and deployability. Enterprises want to use these agents because they are proven and widely adopted, but the hyperscalers' ban revealed that running them directly is too risky. By adding three specific technical layers — isolated environments to prevent cross-contamination, validation of runtime updates before deployment, and standardized tool-connection protocol (MCP) integration — Cloudways is converting an infrastructure problem into a revenue opportunity. The company is not selling the agents themselves; it is selling the trust and operational safety required to run blacklisted code in production.

FAQ

Why would enterprises use agents that Meta, Google, Microsoft, and Amazon banned?
The agents themselves (OpenClaw with over 386,000 GitHub stars and Hermes with over 228,000) have useful capabilities, but the hyperscalers banned them because of instability during operation. OpenClaw had approximately 530 vulnerabilities, including plaintext credential exposure, over 600 malicious skills, and 1.5 million leaked API tokens. Cloudways is solving this by wrapping them in isolated environments and code validation, allowing enterprises to use the agents without the direct risks the hyperscalers encountered.
What happened in February 2026 that triggered the ban?
An OpenClaw agent deleted hundreds of emails from Summer Yue, Director of Alignment at Meta Superintelligence Labs. The failure occurred during context window compaction — a process that compresses long data sequences to fit an AI's memory — and it stripped away critical safety instructions, leading Meta, Google, Microsoft, and Amazon to blacklist the framework.
How much does Cloudways charge and what do customers pay for themselves?
Cloudways charges a monthly hosting fee ranging from $4.99 promotional to $79.99 for standard tiers. Customers use a bring-your-own-key model, meaning they pay Cloudways only for the managed hosting and infrastructure controls, and cover their own LLM (AI model) usage costs separately.
Yahoo Finance AIRead Original Article

Get the latest Large Language Models news every morning

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytime

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleNvidia to backstop OpenAI's $105B Ohio data center

The AI news that matters, in one minute each morning.

Sign up free