On October 27, 2026 Council unanimously passed the SAFE Act. SAFE stands for secure, accountable, fair, and explainable. It sets written standards for how the City buys, uses, and shares data from technology, and it takes effect immediately. I've been working on it for most of this year, including input from Councilmember Rogers and Councilmember Metayer, and refined it with the City Manager, City Attorney, and Chief of Police. Members of the public also weighed in, including a cybersecurity and AI attorney who lives in Pflugerville that volunteered her time to the cause.
Why this exists
The SAFE Act was in progress well before Flock became a headline. I started on it because the City needed rules for high-risk technology before a problem showed up, and that covers far more than police technology. AI tools that influence permits or enforcement, systems that monitor public space, and anything a contractor does with City data on the City's behalf carry the same risks: unclear limits on who sees the data, no written approval, and no public record.
Flock showed what that gap costs. In August, records showed 459 outside agencies had access to Pflugerville's Flock data, and only 80 were authorized, on a system configured to block that sharing. Council voted unanimously on August 25 to terminate the contract. The SAFE Act sets the standard in advance, so the next vendor, pilot, or renewal meets it before the City signs.
How risk classification works
Every technology system the City uses will be classified by risk. Low and moderate risk systems stay with the City Manager. High-risk systems need a public vote by City Council.
A system is high-risk if it is used for law enforcement, investigation, public-space monitoring, networked surveillance, or biometric identification. It is also high-risk if it has standing outside access to City data, or if it influences a consequential decision like enforcement, permits, fines, discipline, or access to services.
Before the City buys, pilots, renews, extends, or materially expands a high-risk system, three things happen:
- Staff writes a review covering legal, privacy, security, operational, accessibility, and fairness risks.
- The City publishes a plain-language summary for the public.
- Council approves the system in a public vote, with the opportunity for public comment.
The approval has to state the purpose, department, categories of data, retention, human review requirements, outside access, and a term.
Classification depends on what a system actually does and its potential for harm, not on its product name, contract structure, vendor description, or configuration. A department also cannot split, relabel, or stage a purchase to avoid approval.
Outside access to City data
This is the section written in response to Flock.
- Standing, recurring, automated, or direct access by another entity requires a written agreement with that specific entity and a Council vote. A shared vendor, platform, or network does not count as that agreement.
- A series of one-off requests cannot be used to build the practical equivalent of standing access.
- Agencies outside CAPCOG (Central Texas) region, and limited-purpose agencies tied to a single campus, facility, or transit system, need a documented Pflugerville connection. Examples are a specific investigation, warrant, or missing person. Membership in the same network or a general claim of regional crime does not qualify.
- Active emergencies can get temporary direct access. It cannot be renewed or chained into standing access, and it appears in the annual report.
- Vendors cannot use City data for their own purposes or for another customer. Under the SAFE Act, that counts as outside access. Vendors also cannot sell the data, train a general-purpose model on it, or build an unrelated product from it, unless Council expressly approves.
When a one-time request for specific information meets the need, the City is directed to use that instead of standing access.
Facial recognition and biometrics
Facial recognition and other biometric identification are prohibited unless Council separately authorizes them after City Attorney review and public comment. The same goes for scraping images or social media for identification, real-time identification of people in a crowd, suspicionless tracking of lawful First Amendment activity, and social scoring.
Established criminal-justice systems of record are carved out, as are non-identifying tools like redaction, object detection, and counting.
Humans stay accountable
AI output cannot be the sole basis for a consequential decision. A trained employee has to understand the system's limits, look at information outside it, and have the authority to reject or override it. Employees remain responsible for work done with these tools.
Residents get a process to report concerns about a system, request human review of an automated action that affects them, and request correction of inaccurate information.
Reporting and transparency
The City Manager must maintain a public technology webpage listing high-risk systems, networked surveillance systems, biometric or identity-search capabilities, and standing outside access arrangements. Each entry describes the system, department, purpose, vendor, categories of data, safeguards, and the Council approval. Information can be withheld only when disclosure would create a security risk, compromise an investigation, or is protected by law.
The City Manager must notify Council within 48 hours of determining that a Material Incident occurred, such as unauthorized outside access or prohibited facial recognition use. A written update follows within 10 business days.
Within 90 days after each fiscal year ends, the City publishes an annual SAFE Act report. For public-safety technology, it has to separate activity, like searches and alerts, from verified outcomes, like leads, cases solved, arrests, and recoveries.
The SAFE Act also treats state law as a floor. It requires compliance with the Texas Responsible Artificial Intelligence Governance Act, Government Code Chapter 2054, and the Department of Information Resources rules, and layers its Council-approval and transparency requirements on top. SAFE goes well beyond state law, though.
What the SAFE Act is not
- It is internal City policy. It does not regulate private businesses, residents, or HOAs.
- It does not create a private cause of action.
- It does not ban police technology. It requires public approval, written safeguards, auditing, and reporting.
- Existing systems continue within their approved purpose and contract term until renewal, extension, or material change.
What happens next
As of October 27, the City cannot newly acquire, pilot, renew, or extend a high-risk system, a networked surveillance system, biometric technology, or standing outside access to City data until review and approval are complete. Routine maintenance, security patches, legally required disclosures, and emergency action for an imminent threat can continue.
Within 90 days, by late January 2027, the City Manager has to stand up the classification process, inventory known high-risk systems and outside access arrangements, and publish the public technology webpage. Full implementation is due within 12 months. Existing contracts get brought into conformity at renewal, extension, or the earliest lawful opportunity.
A SAFE Technology Commission, created by separate ordinance, will review each Council approval and the annual report in an advisory role. Council keeps the final vote.
Axon and Clearview AI are the next systems I want reviewed under this framework, those are expected within the next 90-days.
Questions about this issue?
Reach out directly or keep reading the council updates for the bigger picture.