Key Takeaways:

• Zero trust and attribute-based controls can enable intelligence sharing while protecting sensitive data.
• Secure AI requires trusted data, protection against model poisoning and deliberate data management.
• Effective data strategies must balance public safety, individual privacy, retention and transparency.

Law enforcement leaders see AI’s potential in helping to detect and thwart fast-moving, cross-jurisdictional threats, but they’re understandably cautious about the risks associated with extensive data sharing. Most law enforcement data is sensitive, and the institutional reflex is to ensure its protection for the integrity of the respective agency and/or investigation.

Fortunately, connecting intelligence and safeguarding shared information isn’t an either/or proposition. With the right architecture, they can exist together. The key: making security an enabler rather than an obstacle. That same flexibility should extend to implementation. Connected intelligence capabilities may need to support a long-term modernization effort, an emerging mission need or a time-sensitive public safety event. Agencies should be able to scale people, technology and support to fit the problem at hand while maintaining continuity and security.

The Case for Zero Trust Architecture

For years, agencies faced what appeared to be a binary choice: lock data down and lose its value or open it up and accept the risk. Modern security architecture has changed that proposition.

Zero trust architectures continuously evaluate access rather than relying on implicit trust, while attribute-based controls can provide more granular access based on user, role, data, context and other factors.

Together, these approaches can help agencies share mission-relevant information while maintaining tighter control over sensitive data. The intent of zero trust is simple: instead of trusting anyone inside a perimeter, zero trust verifies every request. Instead of granting blanket access, attribute-based controls release the right data to the right person under the right conditions. As a result, an agency can share what strengthens the mission while keeping everything else protected down to the individual data element.

Managing the Data and the Model

Protection is about more than keeping sensitive information out of the wrong hands. It’s also about guarding against poisoning—deliberate manipulation of training data to distort a model. If attackers can influence what a model learns, they don’t need to breach a database to cause damage. Such threats reframe data security as a mission-integrity issue as well as a privacy issue. They’re also the reason connected systems must be built with defense in mind from the start.

There’s another practical dimension as well. Getting value from AI means ensuring that the right data elements are accessible. That’s a real challenge for the sensitive and proprietary datasets that hold the most value. As agencies work to strengthen their models, they often need to bring their own trusted data together with open-source data. Each stream must be assigned a valuation and accounted for in the agency’s data strategy and management plans. Well-executed programs must be deliberate about what’s worth connecting and what it costs to do so.

Balancing Public Safety and Individual Privacy

Surveillance and data analysis involve collecting and processing personal information, often without explicit consent. That raises challenging questions about how to balance public safety against individual privacy. With agencies now storing petabytes of data, a single breach could expose sensitive information about a large segment of the public. That makes zero trust security architecture an obligation rather than a mere capability.

Data retention compounds the complexity. Policies around how and how long data is stored must evolve to keep pace with the growing volume and complexity of what agencies collect. There won’t always be a one-size-fits-all rule.

Transparency can cut both ways. In some contexts, such as with financial data, full openness can undermine security instead of supporting it. Getting this right is nuanced work that must account for the particulars of each use case, developed and updated with forethought and a willingness to adapt.

Once an agency can connect and share intelligence securely, the final question is whether the decisions that intelligence informs will hold up under scrutiny. That’s the topic we tackle in the final installment of this series.

See how SAIC helps agencies connect data, intelligence and technology to know what matters sooner and act with confidence.

Connected intelligence is one part of a larger story about turning data into trusted intelligence across the mission. See how agencies are putting The Power of Knowing to work.

Explore the Power of Knowing