Fight fear, uncertainty, and doubt about AI data security

Let innovators use technology securely, responsibly, and confidently.

The need for AI-powered innovation couldn’t be clearer

• Attackers are racing to use AI for evil; defenders need it to fight back.

• Companies can save thousands of man-hours spent on repetitive tasks with it.

• The American healthcare system is an expensive joke crying out for automation of low-value processes like note-taking and billing management.

But fear, uncertainty, and doubt (FUD) are holding back AI adoption

• According to Gartner, ~30% of enterprises deploying AI had an associated breach.

• Lucidworks saw a ~3x increase from 2023 to 2024 over AI data security concerns

• In another report it found ensuring security was the top worry for executives.

• Scale AI said that of the 60% respondents to a survey who had yet to adopt AI, security concerns (along with lack of expertise) was a top reason.

• A senior Department of Veterans Affairs official told TechCrunch “significant privacy and security concerns surround using generative AI in healthcare.”

When people talk about FUD in security circles, it’s often with negative connotation. But FUD isn’t necessarily incorrect or unjustified.

Using AI does create risk.

But you can manage it without fear.

Here are three steps to address FUD:

1) Establish - and demand from vendors - clear AI training policies

Concerns about data leakage from models trained on confidential data are understandable.

But companies can:

• Opt-out of training with Software-as-a-Service (SaaS) AI tools.

• Filter user prompts before sending.

• Manage models themselves.

2) Use threat modeling to design secure architectures

Retrieval and similar data pipelines can be a risk if not set up correctly.

Best practices include:

• Controlling data access with rules, not system prompts.

• Minimizing complexity and number of components.

• Implementing guardrails and content filtering.

3) Red-team and test applications before launch

You never know how an AI application will perform until you try it in the real-world.

Have experts:

• Make sure it meets your business requirements.

• Penetration test it to find security gaps.

• Test it for data poisoning.

Need help fighting the FUD?

StackAware lets innovators in healthcare, financial services, and B2B software use AI securely, responsibly, and confidently. Through AI

• risk assessments

• penetration testing

• governance programs

we help customers leverage cutting edge technology without FUD.

Want to learn more about what we do?