ModelsAugust 17, 2026

OpenAI disbands preparedness team responsible for assessing catastrophic ai risks

OpenAI has reportedly dissolved its preparedness team, the group tasked with evaluating and mitigating potential catastrophic risks from advanced AI models. This internal restructuring follows several high profile departures from the company safety and alignment divisions.

Why it matters

For enterprise teams, this shift highlights a changing landscape in how major model providers balance rapid deployment against rigorous safety frameworks. It underscores the importance of internal governance and independent risk assessment when integrating foundational models into production.

Key points

The preparedness team focused on identifying risks related to cybersecurity and chemical or biological threats
Responsibility for safety testing is expected to be integrated across other research and engineering departments
The move comes amid ongoing debate regarding the prioritisation of safety versus commercial speed in AI development

More AI news

Models

Open source tool bypasses Anthropic's new Claude watermarking system

A developer has released an open source tool that strips the invisible watermarks Anthropic recently added to its AI outputs. These watermarks were intended to help the company comply with upcoming EU AI Act transparency requirements. The bypass demonstrates how quickly community driven tools can neutralise provider side safety features.

Models

China's focus on open source AI models reshapes the global development landscape

US trade restrictions aimed at containing China's AI progress have instead accelerated the growth of its open-source ecosystem. Major Chinese tech firms are increasingly releasing powerful open models that compete directly with leading proprietary systems from the West.

Tooling

Faraday AI outperforms major models in scientific research replication

Inherent has launched Faraday AI, a model specifically engineered to replicate complex scientific research with high accuracy. Recent testing indicates that the system exceeds the performance of generalist models from OpenAI and Anthropic in technical validation tasks and data synthesis.

Ready to build something
extraordinary?

15 minutes. No pitch deck. Just a conversation about what AI can do for your team.

Talk directly with our AI specialists

15 min, no strings

No sales pressure

Prototype in 7 days