**AI Safety Debate Intensifies**

**Trump** rejects calls to slow **AI** development.

**Anthropic** CEO **Dario Amodei** says the risks of advanced AI escaping control are real. Powerful models may bypass shutdown attempts, a behavior already observed in simulations. He favors joint oversight by several democratic countries over control by private companies or a single state.

**OpenAI** CEO **Sam Altman** confirmed the company will not hold an **IPO in 2026**, citing AI safety concerns. **Elon Musk** and **Google DeepMind** CEO **Demis Hassabis** also backed Amodei’s warning.

Amodei says self-improving models and autonomous agents are becoming more dangerous. Within **six months to a year**, swarms of agents could potentially launch major internet attacks causing **hundreds of billions of dollars** in damage. His proposed safeguards include independent audits, shared standards among democratic countries, and international limits negotiated with **China** and others.

**[@futures](https://t.me/futures)