The Department of Commerce will start to test new AI capabilities, which could delay the release of new marketing tools that are powered by LLMs.
Big tech firms Google, xAI, and Microsoft have agreed to submit their new AI models for testing through the Commerce Centre for AI Standards and Innovation (CAISI).
The voluntary agreement with the US Department of Commerce will see new capabilities evaluated through “testing, collaborative research, and best practice development related to commercial AI systems”.
The new review process, aimed at protecting US national security, will evaluate the models for cybersecurity, biosecurity, and chemical weapons risks before they are publically available, as well as post-deployment.
“Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,” said CAISI’s director, Chris Fall. “These expanded industry collaborations help us scale our work in the public interest at a critical moment.”
CAISI revealed it has completed over 40 evaluations, with state-of-the-art models remaining unreleased – although it did not specify which models it is referring to.
This comes as AI safety makes headlines, with Anthropic deciding not to release its latest AI model, Claude Mythos, to the public, citing concerns the LLM could find and exploit zero-day flaws (previously unknown and unpatched vulnerabilities in software and hardware), turbocharging cybercriminal capabilities.
The UK has also reached an agreement with Microsoft to test its frontier models to “assess safeguards, and help mitigate national security and large-scale public safety risks”.
“As AI capabilities advance, so too must the rigor of the testing and safeguards that underpin them. We will apply what we learn from these partnerships directly into how we design, test, and deploy AI systems, ensuring that progress in evaluation science translates into safer, more secure products for our customers,” said Natasha Crampton, chief responsible AI officer for Microsoft, in a blog post.
High profile figures such as US senator Edward Markey have raised concerns about LLM safety when it comes to consumer protection, privacy, and user safety.
Guardrails within the models have been incrementally introduced to address these concerns – and as more immediately threatening issues are tackled like cybersecurity and biosecurity, this sets the precedent for stricter testing to evaluate for harmful content, hallucinations, and manipulation.
For marketers, this marks a shift from the ‘wild west’ use of LLMs and AI to a more controlled environment focused on compliance. It could also mean a delay in new marketing tools powered by these models.
