How Privacy Tech Can Make AI Safer for You: What the New GAO Report Says

If you’ve used ChatGPT, Microsoft Copilot, or any other AI tool, you’ve probably wondered what happens to the data you type in. These services are incredibly useful, but they also raise real privacy questions. A new report from the U.S. Government Accountability Office (GAO) released on May 20, 2026, suggests that the answer may lie in something called “privacy-enhancing technologies” (PETs). Here’s what the report found, how these technologies work, and what you can do to protect your own data in the meantime.

What Happened

The GAO report, titled Privacy Technology: Key to Safer AI Adoption, examines the growing use of AI across federal agencies and the privacy risks that come with it. The report notes that while AI can be powerful, it often relies on large amounts of personal data—data that can be exposed, misused, or reidentified if handled poorly. The GAO recommends that agencies adopt PETs to reduce these risks. Specifically, it highlights three categories of privacy tech:

  • Differential privacy – adds mathematical noise to data so that no individual’s information can be singled out.
  • Federated learning – trains AI models on data that stays on your device, only sending model updates (not raw data) to a central server.
  • Homomorphic encryption – allows computation on encrypted data without ever decrypting it, so the service never sees your actual information.

These aren’t new inventions—researchers have been working on them for years. But the GAO’s report is a signal that the U.S. government now sees them as essential infrastructure for safe AI adoption, not just academic curiosities.

Why It Matters

For everyday users, the importance of PETs is straightforward: they make it harder for companies to misuse your data, even if they want to. When an AI service uses differential privacy, for example, it can learn patterns from millions of conversations without learning anything specific about you. Federated learning means an AI keyboard app could improve its predictions without uploading your entire typing history.

That said, most consumer AI tools today do not fully deploy these technologies. ChatGPT, for instance, uses some privacy measures, but it still collects and stores conversation data for training. The GAO report is aimed at government procurement, but it also puts pressure on private companies to follow suit. The practical effect for you: over time, you may see more services advertising that they use “differential privacy” or “federated learning.” Knowing what those terms mean helps you make informed choices.

But the report is not a guarantee. It acknowledges that PETs are not a silver bullet—they can affect model accuracy, slow down responses, or be complex to implement. You should not assume that every tool claiming to use PETs is fully private. The technology is still maturing.

What You Can Do

While companies catch up, you can take a few concrete steps to limit your exposure:

  1. Check privacy policies for specific terms. Look for mentions of “differential privacy,” “federated learning,” “on-device processing,” or “no data retention.” If a policy is vague or says data “may be used for training,” assume your inputs are not private.

  2. Use AI tools that offer a privacy mode. Some services let you disable chat history or opt out of training. For example, ChatGPT has a “Temporary Chat” option. Turn it on for sensitive questions.

  3. Be careful what you share. Avoid typing personally identifiable information (full name, address, phone numbers, health records) into any AI chatbot, even if it claims to be private. PETs reduce risk, but they don’t eliminate it.

  4. Consider local AI options. Apps and models that run entirely on your device (like Apple’s on‑device intelligence or some open‑source language models) never send your data anywhere. This is the strongest privacy guarantee you can get today.

  5. Watch for new features. As agencies and companies adopt the GAO’s recommendations, expect to see more transparency. If a service starts offering “privacy‑enhanced AI” in its settings, read the fine print and test it.

The GAO report is a helpful roadmap, but your own caution remains the best tool. PETs will make AI safer, but they are not here yet at scale—and even when they are, they won’t replace good habits.

Sources

  • U.S. Government Accountability Office, Privacy Technology: Key to Safer AI Adoption, published May 20, 2026.
  • MeriTalk, “GAO: Privacy Tech Could Be Key to Safer AI Adoption,” May 20, 2026.