MOBILITY ENGINEERING Produced by SAE Media Group
MOBILITY ENGINEERING
Magazines
Magazine cover

Current Issue

Archives

Magazine cover

Current Issue

Archives

Magazine cover

Current Issue

Archives

Magazine cover

Current Issue

Archives

Open Safety is the Shortcut to Safer ADAS/AD

Treat foundational AV safety like seatbelts – make it non-proprietary and universal. An open safety stack, shared scenarios, benchmarks, and core validation tools can speed certification, reduce duplicated V&V and build public trust while preserving vendor differentiation.

An Open Safety stack includes three main categories: base infrastructure, a shared validation layer and a differentiation layer. (DEEPEN AI)

The bottleneck isn’t compute - it’s verification.

Autonomous features are shipping in more vehicles and markets, but the gating factor is no longer raw compute. It's whether developers and regulators can verify systems against requirements and validate them against real-world operating design domains (ODDs) with confidence and repeatability. Today, many safety-critical components, from scenario libraries to pass/fail criteria, live in proprietary silos. That fragmentation slows regression testing, complicates regulator audits across regions, and duplicates effort across the industry. The result is an expensive, bespoke path to certification for every program and geography.

2025 and 2026 serve as an inflection point. As OEMs expand Level 2+ to Level 3 features and explore limited-ODD automated driving, the cost and time to assemble defensible safety evidence are rising. Without shared artifacts, every company rediscovers the same edge cases, invents parallel benchmarks, and negotiates one-off interpretations with regulators. We, the AV industry, can do better, but we must do it together.

The seatbelt lesson: make the baseline universal.

In 1959, Volvo engineer Nils Bohlin introduced the three-point seatbelt. Volvo opened the patent so any automaker could implement it. Universal adoption followed, and the design is credited with saving over a million lives. The lesson is not that competition should stop; it’s that baseline safety should be treated as public infrastructure. When the tide rises for everyone, innovation moves where it belongs, above the baseline.

Translating that to software: safety-critical validation artifacts, not your secret sauce in perception or planning, should be shared industry-wide. When core, non-differentiating safety elements are open and common, we certify faster, learn faster, and build trust faster.

What an “open safety stack” includes

Shared validation layer (open by default):

  • Scenario libraries that capture edge cases and rare events with clear semantics and metadata.
  • Benchmarks and metrics with published pass/fail criteria aligned to stated ODDs.
  • Test harnesses and reference adapters that let developers reproduce results across tools.
  • Audit artifacts (coverage, traces, rationales) that are machine-readable and regulator-friendly.

Where vendors still differentiate (what’s kept proprietary):

  • Perception and planning models, fusion strategies, and training data pipelines.
  • Integration choices, hardware abstraction, software performance optimizations.
  • Human-machine interface (HMI) and user experience.
  • Data strategy (collection, curation, active learning) and fleet ops.

A better outcome

With the safety baseline shared, the industry learns from incidents collectively, reuses validated tests, and produces certification packets that are traceable, comparable, and easier for regulators to review.

In fact, early, practical steps in this direction have already been made. Initiatives that make safety quantifiable and shareable, through public scenario repositories, transparent benchmarking, and certification-oriented tooling, demonstrate how common artifacts can align OEMs, suppliers, researchers, and regulators. Similarly, open-source AD stacks and kits show how a neutral, community-governed substrate can speed interoperability and reduce the cost of adoption for new entrants. These are not attempts to commoditize innovation; they’re attempts to open-source safety, so differentiation can focus on product experience and performance rather than rebuilding the same safety scaffolding in private.

An engineer’s view: how verification and validation changes daily.

Engineers care about what lands on their desk. Here’s how an open safety stack reshapes the daily workflow:

  1. Scenario-to-simulation. A shared scenario (with ODD tags, actor behaviors, weather/lighting permutations, and acceptance criteria) runs in multiple simulators via a reference harness. The same JSON/YAML definition triggers the same behaviors everywhere.

  2. Regression at scale. Vendors add program-specific variants, but shared core suites mean each software change is tested against community-accepted hazards first. Failures are recorded with standard traces and deltas.

  3. Coverage you can reason about. Coverage metrics (e.g., by scenario class, environmental conditions, map semantics) use a shared taxonomy. Engineers no longer translate private labels into regulator-friendly language, because the labels are regulator-friendly.

  4. Certification packet assembly. Evidence exports are standardized: scenario IDs, metric outcomes, justification notes, traceability to requirements, and links to replay artifacts. Auditors see familiar structure regardless of vendor.

  5. Feedback to the commons. When a rare field event emerges, a minimal, privacy-safe scenario is contributed back to the shared set with reproducer code and acceptance tests.

The entire ecosystem benefits quickly from this.

The practical wins are substantial: fewer bespoke test frameworks, faster onboarding of partners and suppliers, and less friction moving between markets with different regulatory authorities.

Policy and ecosystem: a rising floor, not a ceiling

Open safety artifacts help standards bodies and regulators in two ways. First, they reduce ambiguity: clear scenario semantics and metrics make guidance more testable. Second, they lower the review burden: when formats and proofs are consistent, audits go faster. None of this prevents an OEM from exceeding the baseline; in fact, it enables higher ceilings by clearing repetitive hurdles at the floor.

For policymakers, transparent artifacts make it easier to maintain living guidance: when the long tail of scenarios is community-curated, updates can be adopted by all participants without restarting from scratch. For suppliers, common definitions reduce integration debt across programs. For startups, shared safety layers remove a heavy upfront cost, encouraging responsible innovation.

Implementation notes (for teams getting started)

  1. Pick a canonical schema for scenarios and commit to converters. Perfect is the enemy of progress. Instead, start with a subset and iterate.
  2. Publish metrics with tolerances. A metric that others can implement identically is more valuable than a perfect one no one can reproduce.
  3. Automate evidence export. Treat the certification packet as a first-class build artifact with CI/CD gates.
  4. Govern lightly, version ruthlessly. Use semantic versioning and require migration notes; instability kills reuse.
  5. Design for regulator empathy. Write acceptance criteria and traces like someone outside your company will read them; because chances are they will.

Addressing common concerns

“We’ll lose our edge.” Your edge isn’t a pedestrian-crossing scenario ID. It’s your perception/planning stack, your data, your integration craft, and your UX. Keep those proprietary and move faster by not rebuilding the non-differentiating safety layer alone.

“Open artifacts will be gamed.” Make gaming visible: publish both metrics and counter-metrics (e.g., robustness checks, distribution shifts). Bad behavior becomes detectable and fixable for everyone.

“Regulators won’t accept open material.” Regulators accept evidence. Open artifacts make evidence consistent, which isn’t only welcomed, it’s preferred, especially when formats are co-developed with standards bodies and industry.

Conclusion: raise the floor, free the ceiling

The seatbelt’s success was not just its design; it was the decision to open it. ADAS/AD needs the same mindset. By treating foundational safety artifacts as shared infrastructure, we shorten certification cycles, reduce duplicated V&V, and accelerate learning from the rare events that matter most. Openness at the baseline is not charity; it’s the most efficient way to scale safer autonomy and the fastest path to public trust.

Mohammad Musa is CEO of Deepen AI. Muhammad Zain Khawaja works on open-source autonomous driving stacks and ecosystem standards for the Autoware Foundation. They wrote this article for SAE Media Group.

Topics:
ADAS Automotive Automotive Manufacturing & Prototyping Simulation Software

More From SAE Media Group

    Magazine cover
    Automotive Engineering Magazine

    This article first appeared in the February, 2026 issue of Automotive Engineering Magazine (Vol. 13 No. 1).

    Read more articles from the archives here.

    Top Stories

    INSIDERUnmanned Systems

    Feature Image Airbus, Boeing Reveal New Autonomous Aircraft Technologies at ILA Berlin 2026

    INSIDERWeapons Systems

    Feature Image A Joint Laser Weapon System for Next Generation Drone and Cruise Missile Defense

    NewsElectronics & Computers

    Feature Image Mercedes-AMG Unveils Racing-Inspired EV

    NewsPower

    Feature Image We Ride Along in Slate's $25,000 Electric Pickup

    Q&AManned Systems

    Feature Image Volvo Turns Attention to Vocational Trucks

    NewsEnergy

    Feature Image GM Energy Bets on Sodium Batteries for Energy Storage

    Webcasts

    Similar Stories

    ArticlesSensors/Data Acquisition

    Feature Image StreetDrone: Rolling R&D Vehicle for Small-Scale AV Developers