Recent events have reinforced a foundational truth of modern data security: data privacy is only as strong as the systems that control access to encryption keys.

A clear example came to light in January, when reporting revealed that Microsoft provided the FBI with BitLocker recovery keys, allowing law enforcement to unlock data that users believed was fully encrypted and private. The keys were accessible because users had agreed to automatically back it up to the cloud provider’s environment by default – a design choice optimized for convenience, not customer sovereignty.

This moment is particularly relevant during International Data Privacy Week, a global reminder that protecting sensitive data is not just a regulatory obligation, but a trust imperative. As organizations reflect on how they collect, access, and protect data an important question emerges:

Who controls the keys to your most valuable data assets?

Convenience vs. control

The Microsoft BitLocker case is not about wrongdoing or misconfiguration. It is about architectural control.

BitLocker encryption worked exactly as designed. However, because they stored recovery keys in a provider-managed cloud environment, Microsoft was technically and legally able to comply with a warrant by providing access to those keys. As reported by Forbes, this meant the FBI could unlock encrypted data without the customer’s direct involvement.

This illustrates a broader issue affecting many modern platforms:

• Encryption keys may be automatically escrowed, without explicit customer intent
• Providers can be compelled to provide third-party access car, not the data owner
• Compliance workflows may prioritize provider obligations over customer control

For enterprises, the takeaway is simple: if your cloud provider can access your encryption keys, your data is never fully under your control.

With AI data security, expanded access equals expanded risk

Encryption keys centralized within provider-managed infrastructure, expand the blast radius of access dramatically. Even when strong safeguards exist, cloud-stored keys introduce additional access paths that organizations do not directly govern.

This risk becomes even more pronounced in environments where:

• AI agents routinely access sensitive datasets
• Machine identities outnumber human users
• Enterprise AI security depends on precise, enforceable access boundaries

Without strong AI data security, AI access control, and AI governance security, organizations risk unintentionally extending trust to systems and providers that were never meant to hold it.

Customer-controlled key models are becoming the standard

It is imperative that forward-looking security programs treat encryption keys as true crown jewels. Rather than defaulting to provider-managed models, such organizations are prioritizing architectures that preserve control by design:

• Bring Your Own Key (BYOK): Customers manage keys while platforms enforce policy
• Hold Your Own Key (HYOK): Providers cannot access keys at all
• Client-side encryption: Keys never leave customer-controlled environments

These models ensure that even under subpoena or legal request, third parties cannot decrypt sensitive data without explicit customer participation – a critical requirement for modern AI risk management, regulatory compliance, and audit readiness.

Trust is an architectural decision

The Microsoft BitLocker case serves as a market-wide learning moment:
Trust is not a feature; it is an architectural decision.

Organizations that prioritize data privacy, governance, and enterprise resilience are moving away from silent defaults and toward intentional control models. They demand transparency, enforce least privilege access, and retain authority over who, or what, can unlock their data.

As International Data Privacy Week reminds us, enterprises cannot define privacy by encryption alone. It is defined by who controls access when it matters most. Learn more about how OpenText can help protect your crucial data.

The post AI data security starts with controlling access to your crown jewels appeared first on OpenText Blogs.