The DHS SAFETY Act in AI Security: Designation, Certification, and What It Actually Means

The DHS SAFETY Act is the single most cited credential in AI physical security procurement, and the one most often misunderstood at the buying table. Vendors say "SAFETY Act Designated" the way software vendors say "SOC 2 Type II"; security directors and risk managers nod, sign, and assume parity. There is no parity. The SAFETY Act has three statutory tiers, two of them confer materially different liability protection, and the gap between a Developmental Testing & Evaluation Designation and a full Designation is wider than most procurement teams realize.

This market analysis decodes the SAFETY Act for security buyers evaluating AI-powered detection platforms. It explains the QATT, Designated, and Certified distinction in the language Congress wrote it; maps which vendors hold which tier as of Q2 2026; and surfaces the contractual and insurance questions a buyer should ask before signing a master service agreement. The objective is not vendor advocacy. It is to give risk-rated decision-makers a defensible reference for the SAFETY Act question that increasingly determines whether AI security technology gets through legal review.

LIVE CAM-12 · INTERIOR CORRIDOR

What the SAFETY Act actually is, in the language Congress wrote it

The Support Anti-Terrorism by Fostering Effective Technologies Act of 2002, the SAFETY Act, was enacted as Subtitle G of Title VIII of the Homeland Security Act of 2002, Public Law 107-296. The statute is implemented through the Code of Federal Regulations at 6 CFR Part 25 and administered by the DHS Science & Technology Directorate's Office of SAFETY Act Implementation (OSAI). Its core purpose, written into the text of the statute, is to ensure that the threat of post-event tort liability does not deter manufacturers and sellers of effective anti-terrorism technologies from developing and commercializing them.

The mechanism is liability management, not technical certification. Anti-terrorism technologies that pass DHS review become Qualified Anti-Terrorism Technologies, or QATTs. A QATT designation triggers a defined set of statutory benefits that limit the seller's exposure if the technology is in use during a declared act of terrorism. Those benefits compound through three tiers, and the difference between the tiers is the difference between a discount on liability and an effective immunity.

The SAFETY Act is therefore a procurement document as much as a security one. A general counsel reviewing an AI security MSA wants to know what statutory floor the vendor sits on. A risk manager pricing terror exposure into the property and casualty book wants to know whether the technology shifts the loss-cost calculation. A federal contractor preparing a bid wants to know whether the platform deployed under the contract will trigger the Government Contractor Defense for the Department's own contracting officers. The SAFETY Act answers all three questions, but only at the right tier.

The three SAFETY Act tiers, decoded

OSAI grants three distinct levels of SAFETY Act protection. Each provides a different statutory benefit, requires a different level of operational evidence, and runs on a different term clock. Treating them as interchangeable on a procurement scorecard is the single most common buyer error.

The shape of the protection looks the same on a marketing slide. It is not the same in a deposition. A Designated Technology compresses the liability tail materially. A Certified Technology effectively closes it through the Government Contractor Defense presumption, which under SAFETY Act jurisprudence can only be overcome by evidence of "knowing and deliberate intent to deceive the government" during the application process. A DT&E Designation provides real but provisional cover, scoped to the trial deployment and bounded by a clock the seller cannot control. Treating these as a single credential is a procurement failure mode.

Why the SAFETY Act tier matters more in 2026 than it did in 2022

The SAFETY Act has been on the books since 2002, but its weight in security procurement has grown sharply over the past three years for three converging reasons. Each reason puts more pressure on the procurement question, and each rewards buyers who know how to read the tiers.

First: state legislation is naming SAFETY Act status as evaluation criteria. A growing number of state-level school security and hospital security mandates either explicitly reference SAFETY Act Designation or use language that effectively requires it. Tennessee's school weapons-detection grant programs, several state hospital workplace violence statutes, and federal grant programs administered through CISA and FEMA all reference SAFETY Act status as a procurement screen. A live tracker of these mandates is maintained in the Q2 2026 state AI security legislation tracker and the broader grant funding resource.

Second: insurance carriers are using SAFETY Act status as a risk-rating input. Property and casualty underwriters and active-assailant policy writers increasingly factor whether a deployed security platform is SAFETY Act protected into the risk-rating model. A buyer running a Designated platform sits in a different risk pool than a buyer running an unprotected platform. The premium impact is small per individual policy but compounds across portfolios.

Third: tort exposure for security failures has expanded. Plaintiff-bar cases following high-profile attacks have established that security technology vendors and their customers can both be named in negligent-deployment, failure-to-warn, and failure-to-detect claims. The SAFETY Act's Designation tier exists precisely to compress this exposure. Where it applies, the limitations on damages, joint and several liability, and forum selection materially change the litigation calculus. The dynamic is most visible in environments where the underlying threat model includes mass-attack scenarios, which is the same reason buyers in K-12 education, healthcare, and houses of worship are now framing the SAFETY Act question early in evaluation. The 2026 Houses of Worship Security Sector Playbook provides a practical deployment and NSGP grant strategy framework for faith communities evaluating SAFETY Act-designated platforms.

This is upstream of the technology question. A buyer evaluating AI threat detection for a school district, a hospital system, or a critical-infrastructure facility is now asking the SAFETY Act question before the false-positive-rate question. That order of operations is a 2025 to 2026 development, not a 2002 one.

The SAFETY Act vendor landscape in AI physical security as of Q2 2026

The DHS Approved Technologies database is the authoritative source for vendor SAFETY Act status. The database is publicly searchable at safetyact.gov/at/, and the entries are the controlling record. A buyer should always confirm the listing rather than rely on a vendor's marketing claim.

As of Q2 2026, the AI-powered weapons and threat detection segment shows a small number of vendors at full Designation, a handful at DT&E status, and a long tail of platforms with no SAFETY Act presence. The table below summarizes the structural picture, anchored to the safetyact.gov database. Vendors are listed alphabetically; the absence of a vendor name indicates no SAFETY Act listing as of the verification date.

Several characteristics of the database matter when reading this picture:

The database is granular at the product level, not the company level. A vendor with multiple products may have one Designated product and others without protection. Buyers should verify the specific platform configuration they intend to deploy, not the company name.

DT&E status precedes full Designation in most cases. ZeroEyes and Omnilert both moved through DT&E before earning full Designation. The Designation step is a meaningful upgrade in liability protection and an indication that the technology has accumulated enough operational data for OSAI to grant the longer term.

Certification, Tier 3, is rare in the AI physical security segment. No widely deployed AI weapons detection platform held Certification status as of Q2 2026 verification. Certification requires a higher evidentiary bar and typically follows a multi-year Designation track record. Buyers asking "are you Certified?" are in most cases asking for a tier no AI security vendor has yet reached, and the more useful question is which Designated tier the platform sits at and when the term expires.

The contractual questions a SAFETY Act-aware buyer should ask

If a deployed security platform might be in service during a declared act of terrorism, the SAFETY Act controls a large portion of the resulting liability allocation. A procurement team that understands the structure can pull material concessions from a vendor and reduce the buyer's own exposure in three contracting dimensions.

These are not exotic clauses. They are the SAFETY Act-aware version of insurance and configuration management language that already exists in well-drafted security technology contracts. The legal value is the difference between a contract that preserves the statutory protection through deployment and one that inadvertently narrows it.

Three false equivalences buyers should reject

The SAFETY Act question often gets compressed in procurement scoring in ways that flatten material differences. Three false equivalences come up most often, and each is worth pushing back on in evaluation.

"DT&E and Designated are basically the same." They are not. DT&E is provisional, capped at 36 months, and terminable at-will by the Department. The protection scope is also limited to acts during the DT&E period itself. A buyer signing a five-year contract with a vendor at DT&E status assumes substantial renewal risk that a Designated vendor's contract does not carry.

"Designated and Certified are basically the same." They are not. Certification adds the rebuttable presumption of the Government Contractor Defense, which is a categorically different liability outcome. The presumption can be overcome only by evidence of "knowing and deliberate intent to deceive the government" during application. In a post-event suit, that evidentiary bar is extraordinarily high. Designation does not include the presumption.

"SAFETY Act protection covers any incident." It does not. The SAFETY Act's liability framework activates upon a Secretary of Homeland Security declaration that an event was an act of terrorism under section 865 of the Homeland Security Act. Many security incidents, including workplace violence, suicide, accidental discharge, and fall events, will not trigger that declaration. SAFETY Act protection is most consequential for the specific class of mass-attack scenarios the statute was written to address. It is not a general litigation shield.

A clear-eyed reading is that the SAFETY Act is a powerful but bounded protection, scoped to a specific event class, layered through three tiers with material differences, and useful as a procurement screen primarily where the deployment context puts terrorism-class events on the realistic threat surface.

How SAFETY Act status fits into a defensible AI security buying decision

SAFETY Act status is a meaningful screen, but it is one input in a broader buying calculus. A platform with full Designation that does not meet the operational threshold for the buyer's environment is still a poor purchase. A platform without a SAFETY Act listing that delivers the right detection accuracy in the right deployment may still be the right answer for a buyer whose threat surface does not credibly include declared acts of terrorism.

The question is therefore how SAFETY Act status fits into the larger framework rather than whether it is the decisive criterion in isolation. Three principles operationalize the screen well:

Use SAFETY Act status as a tier-discriminator at the long-list stage. If the buyer's threat surface includes mass-attack scenarios, which is the case for most schools, hospitals, houses of worship, large public-facing commercial buildings, and critical infrastructure, vendors at Designated or higher status sit in a different risk pool than vendors without. Use the tier to narrow the long list before deeper technical evaluation.

Use technical performance as the decisive screen at the short-list stage. Once the SAFETY Act floor is established, the decision moves to detection accuracy, false-positive behavior, integration depth, alert routing, and the operational realities covered in the 2026 definitive guide to proactive computer vision and the failure-mode analysis in the perimeter intrusion 90-second window report. SAFETY Act status does not certify performance; it certifies eligibility for liability protection. How the platform works and what the detection modalities actually catch is the operational evaluation.

Use contracting language to preserve the protection through the term. The SAFETY Act benefits flow only when the technology is deployed in the configuration covered by the Designation. Procurement language that locks the configuration scope, requires renewal commitment from the vendor, and pulls the OSAI letter into the contract is the difference between a SAFETY Act platform on paper and one that actually performs in a litigation event.

Frequently asked questions about the SAFETY Act in AI security procurement

Continue the research

This market analysis covers the SAFETY Act tier framework and the Q2 2026 vendor landscape. For deeper reading on specific pieces of the procurement and deployment picture:

• 2026 definitive guide to AI-powered threat detection, technical reference for the underlying detection technology, including the architectural principles OSAI evaluates during a SAFETY Act review.
• Q2 2026 state AI security legislation tracker, the regulatory mandates that increasingly reference SAFETY Act status as a procurement screen.
• Autonomous Security: how agentic AI is replacing alert-first architectures, the next-generation operational model that sits on top of Designated detection platforms.
• AI gun detection solution page, the technical detail behind the modality most often referenced in SAFETY Act applications.
• Security legislation tracker, the actively maintained database of state and federal mandates relevant to security technology procurement.