---
title: 'In the News: September 3, 2026'
description: "Anthropic replaces its enterprise data-retention mandate with customer-held storage, and Cursor lets cloud agents run inside a team's own infrastructure."
canonical_url: 'https://darkfactory.dev/news/2026-09-03-evening'
markdown_url: 'https://darkfactory.dev/news/2026-09-03-evening.md'
collection: news
date_published: '2026-09-03T19:15:00-04:00'
date_modified: '2026-09-03T19:15:00-04:00'
---

# In the News: September 3, 2026


Anthropic announced Enterprise Frontier Safeguards on September 1, replacing the data-retention rule it added in June. A day later, Cursor gave cloud agents a way to execute entirely inside a customer's own infrastructure. Both moves hand more control over agent execution and data back to the people running it.

## 1. Anthropic replaces its data-retention mandate with a customer-held alternative

**[Developing Enterprise Frontier Safeguards with our customers](https://www.anthropic.com/news/enterprise-frontier-safeguards)** · Anthropic, Sep 1, 2026

Anthropic is rolling out Enterprise Frontier Safeguards (EFS) to replace the 30-day data retention policy it attached to Fable 5 in June. "EFS works by storing data in cloud infrastructure controlled by the customer, not Anthropic," the company wrote, with automated misuse monitoring running on that data instead of a human review team at Anthropic. Flags for suspected credential theft or attempts to develop offensive cyber or biological capabilities go straight to the customer's own security team, not Anthropic's. Anthropic says it built EFS with more than 100 customers, including the CISOs of Goldman Sachs, Morgan Stanley, Citi, Bank of America, and Wells Fargo through the Analysis and Resilience Center for Systemic Risk, and that it will not charge for the feature. EFS covers Claude Code, Claude Enterprise, the Claude Platform, Amazon Bedrock, Google's Agent Platform, and Microsoft Foundry, rolling out in phases starting this fall.

**Why it matters:** The June retention rule was Anthropic's answer to misuse that spans multiple sessions and accounts, the kind an agent operating autonomously can commit over days rather than in one request. Regulated customers pushed back hard enough that Anthropic redesigned the enforcement mechanism around customer-held storage rather than dropping it. Anyone running Claude Code inside a bank, hospital, or law firm now has a path to frontier models without also handing Anthropic long-term custody of session logs.

## 2. Cursor lets cloud agents execute inside a team's own infrastructure

**[Self-hosted machines](https://cursor.com/changelog/self-hosted-machines)** · Cursor (Anysphere), Sep 2, 2026

Cursor's cloud agents can now run tool calls on machines a team owns rather than on Cursor's own infrastructure. "Your codebase, build outputs, and secrets all stay on internal machines running in your infrastructure, while the agent handles tool calls locally," Cursor wrote. A single laptop or VM can register as "My Machines" for personal use, or a team can register a named worker pool that scales up as requests arrive and hibernates idle machines to save cost. Cursor lists integrations for running self-hosted agents on AWS Lambda, Coder, Cloudflare, Daytona, Modal, Namespace, Vercel, and E2B, and self-hosted workers now support computer use (click, type, take screenshots, drive a browser) on Linux and Mac.

**Why it matters:** Cursor gives teams a managed agent interface and dynamic worker scaling while keeping secrets and build artifacts on infrastructure they control. Anthropic's EFS addresses the same control problem at the model layer; Cursor addresses it at the execution layer.