---
title: "AI Safety Warnings and Agent Control Risks | RinggCity"
description: "RinggCity Edition 52 covers AI safety warnings, PCOS testing, OpenAI model behaviour, Ringg MCP, Astra for Law and enterprise agent control."
canonical_url: "https://www.ringg.ai/newsletters/ai-safety-agent-control"
last_updated: "2026-09-18T10:49:11.000Z"
---

# Is It Time to Slow Down AI?

RinggCity 52nd Edition | Published on 18 Sep 2026

Subscribe and join 10k+ builders in tech

## This week in AI

### 1 . Are AI leaders starting to fear the technology they built?

Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman are backing a slower, more coordinated approach to frontier AI development, with stronger testing and independent safety checks.

### 2 . Can AI make PCOS testing simpler?

Hyderabad researchers have received a ₹4.98 crore ICMR grant to develop an AI-powered blood test for PCOS. The project aims to identify biomarkers that could make PCOS diagnosis more accessible

### 3 . Why are OpenAI models behaving unexpectedly?

OpenAI has disclosed six cases of models behaving unexpectedly, including attempts to evade oversight and act without authorisation.

## Our take

### Is It Time to Slow Down AI?

More than a decade after Stephen Hawking warned that advanced AI could threaten humanity, several tech leaders are raising similar concerns. Amodei, Altman and Elon Musk are now calling for more caution as AI systems become increasingly autonomous.

There's an interesting tension here. The same capabilities that make AI useful enough to build businesses around also make it harder to predict how these systems might behave in complex environments. For companies deploying AI agents, this makes control, testing and clear boundaries as important as capability. The next phase of AI adoption may depend as much on knowing what an agent should not do as on what it can do.

## What's Ringging

### Introducing Ringg MCP

Ringg is now connected to the AI tools you already use. With the Ringg AI MCP Server, you can use Ringg with ChatGPT, Claude, Cursor, Claude Code, and other MCP-compatible clients to ask questions, place calls, read transcripts, create assistants, and analyse performance, all in natural language.

## AI product & startup updates

OpenAI takes AI into legal work

OpenAI launched Astra for Law, a version of GPT-6 Astra built for legal research, drafting and custom legal applications. It is being piloted with major law firms and integrates with platforms including Harvey, Relativity, Clio and Thomson Reuters.

*   **Anthropic** launched **Claude for Financial Advisors**, integrating with tools from **BlackRock**, **Charles Schwab** and **Addepar** for portfolio reviews, client preparation and follow-ups.
*   **Factory** raised **$200M** at a **$5B valuation**, tripling its valuation as demand grows for **AI agents** that handle software engineering tasks.
*   **Crusoe** raised **$3.9B** at a **$30.9B valuation**, as demand for **AI-focused data centres** and compute infrastructure continues to surge.
*   **Delos Data** raised **$100M** to build chips and software for faster data movement inside **AI data centres**.

## Footnotes

The AI race may need a speed limit.

Good luck explaining that to Silicon Valley.

![Edition - 52 - footnotes](https://images.prismic.io/ringg-ai/kmGQl52nBBIuZi6W_Funny-AI-Memes_GettyImages_9_1.jpg?auto=format%2Ccompress&fit=max&w=1920)

## More from Ringg City

[

### The G20's AI Regulation Problem

g20 ai regulation problem | 04 Sep 2026

](https://www.ringg.ai/newsletters/g20-ai-regulation-problem)[

### The AI Infrastructure Bill Is Getting Bigger

ai infrastructure debt financing | 24 Aug 2026

](https://www.ringg.ai/newsletters/ai-infrastructure-debt-financing)[

### Second Time's the Alarm

anthropic ai agent safety scare | 18 Aug 2026

](https://www.ringg.ai/newsletters/anthropic-ai-agent-safety-scare)

Book a Demo

Ready to deploy production-grade voice AI?

See how Ringg AI helps teams launch reliable voice agents across support, sales, operations, and industry workflows.

[Book a Demo](https://www.ringg.ai/book-a-demo)

Source: https://www.ringg.ai/newsletters/ai-safety-agent-control