OpenAI recently launched a new feature called “company knowledge,” which promises to boost productivity by giving businesses deeper access to their own data. It’s designed to pull information from apps like Slack, Google Drive, SharePoint, and even GitHub, then use AI to give quick, accurate responses. This could mean faster meetings, better decision-making, and a more seamless workflow. But with great power comes serious concerns about privacy, security, and trust.

What is OpenAI’s Company Knowledge?

This new tool is built on a version of GPT-5 that’s trained to look across multiple sources of enterprise data. OpenAI claims it can create detailed briefings before meetings, pulling in recent messages, emails, notes, and support tickets. Every response is supposed to include sources so users know where the information came from—kind of like having a transparent AI assistant. The idea is to make work easier by having AI bring together all relevant information in one place.

How Will Data Be Used and Protected?

OpenAI says the data used for company knowledge will respect existing permissions. That means the AI can only access what a user is authorized to see. However, the company hasn’t clearly explained what it might do with the data beyond that. Industry experts worry that sensitive information could end up in the wrong hands or be used to train future AI models. Because the data could be accessible by many employees, contractors, or even third parties, tracking its exact path is difficult. If sensitive info leaks outside the company, proving where it came from can be nearly impossible.

Trust and Adoption Challenges

Many enterprise IT leaders are wary of trusting a relatively young company like OpenAI with such deep access. They’re concerned about privacy, data security, compliance, and the risk of vendor lock-in. Some see the potential benefits—like saving time and increasing efficiency—but worry that the risks might outweigh the gains. As one expert explains, companies are increasingly moving toward connected AI systems that integrate with their existing tech stacks. This makes data security and governance even more critical, since these solutions often involve high-risk data sharing.

OpenAI’s statement emphasizes that the AI will only access data that users already have permission to view. Still, many industry insiders believe that relying solely on legal contracts and permissions isn’t enough. Once data leaves the company’s internal systems, it becomes harder to control. If sensitive info appears on the dark web or in the hands of malicious actors, tracing its origin becomes nearly impossible. This raises serious questions about how prepared companies are to govern employee use of AI tools and protect their data.

Risks and Future Considerations

Experts warn that unintentional data leaks are a real concern. Without proper audit trails, it could be hard to tell if privileged info was accessed or shared unintentionally. Some industry leaders believe that unless organizations build dedicated, secure instances of AI models—possibly requiring their own data centers—they’re taking a big risk by deploying these tools broadly. For many companies, especially those with sensitive data, the dangers outweigh the potential productivity benefits.

Cybersecurity professionals also point out that many organizations lack the strict data classification controls needed to safely use AI at this level. For those that do have strong controls, there might be some benefit, but only if they manage access tightly and understand exactly how data flows through these systems. As one expert puts it, the main question is: how long does access last, and how is that access revoked once it’s no longer needed?

In the end, OpenAI’s new enterprise feature is full of promise but also fraught with risks. Companies must weigh the potential productivity gains against the serious security and privacy challenges. Trust in the provider, clear policies, and robust safeguards will be key before adopting such powerful tools at scale.

Inspired by

• https://www.computerworld.com/article/4079011/openais-company-knowledge-wants-access-to-all-of-your-internal-data.html

Sources

• gmpg.org
• forrester.com
• openai.com
• marketri.com
• procircular.com