Microsoft says its AI security tool not only finds bugs, but also fixes them

Microsoft is rolling out a up-to-date feature called “correction” that builds on the company’s efforts to combat AI inaccuracies. Customers who utilize Microsoft Azure to power their AI systems can now benefit from the ability to automatically detect and rewrite incorrect content in AI results.

The remediation feature is available in preview as part of Azure AI Studio, a set of security tools designed to detect vulnerabilities, find “hallucinations,” and block malicious prompts. Once enabled, the remediation system will scan and identify inconsistencies in AI results by comparing them to the client’s source material.

It will then highlight the error, provide information about why it’s incorrect, and rewrite the content in question—all “before the user can see” the inaccuracy. While this seems like a helpful way to address the nonsense often spewed by AI models, it may not be a completely foolproof solution.

Vertex AI, Google’s cloud platform for companies developing AI systems, has a feature that “grounds” AI models by checking results against Google Search, the company’s own data, and soon, third-party datasets.

In a statement for TechCrunchMicrosoft spokesperson said the “correction” system uses “small language models and large language models to match results to grounding documents,” meaning it is also not immune to errors. “It is important to note that grounding detection does not solve the problem of ‘accuracy,’ but it does help match generative AI results to grounding documents,” Microsoft said TechCrunch.