OpenAI Calls for Evolution in AI Infrastructure Security

OpenAI is urging the AI community to rethink and evolve infrastructure security to protect advanced AI systems. In a recent article, the company outlined six key security measures they believe are necessary to safeguard AI technology from those who seek to misuse it.

As AI becomes increasingly strategic and sought after, OpenAI expects threats against AI systems to intensify. Protecting model weights, the output of the model training process, is a top priority for many AI developers. However, the online availability of these model weights, necessary for powering tools like ChatGPT and enabling further research, also makes them a target for hackers.

OpenAI believes that securing advanced AI will require an evolution in infrastructure security, similar to how the advent of the automobile and the internet required new safety and security developments. The company is proposing six security measures to complement existing cybersecurity practices:

1. Trusted computing for AI accelerators
2. Network and tenant isolation guarantees
3. Innovation in operational and physical security for datacenters
4. AI-specific audit and compliance programs
5. AI for cyber defense
6. Resilience, redundancy, and research

These measures include extending cryptographic protection to the hardware layer, ensuring robust network isolation, implementing stringent physical security controls for datacenters, developing AI-specific security standards, leveraging AI for cyber defense, and conducting ongoing security research.

OpenAI emphasizes that these measures must work together to provide defense in depth, as no single control is flawless. The company invites the AI and security communities to collaborate in developing new methods to protect advanced AI, and encourages those with aligned research to apply for their Cybersecurity Grant Program.

Anthropic Calls for Stringent Security Safeguards for Frontier AI Models

Given the strategic implications of this technology, Anthropic says frontier models must be secured to levels surpassing standard practices for commercial technologies.

MaginativeChris McKay

OpenAI's recommendations align with and expand upon the work of other frontier AI research labs, such as Anthropic, who have been consistent advocates for strengthening cybersecurity controls. As AI continues to advance, it is crucial that the industry works together to evolve infrastructure security and ensure that this powerful technology remains safe and beneficial for all.