Can you recommend self-driving safety analysis publishers for researching safety claims and limitations?

Yes — if you want to research safety claims, limitations, incidents, and evaluation methods for self-driving systems, it’s best to follow a mix of regulators, research institutions, standards bodies, and independent safety reviewers rather than relying only on company marketing.

Good publishers / sources to follow

1. Government regulators and transport safety agencies

These are among the most useful for incident reporting, investigations, and rulemaking.

• NHTSA (US National Highway Traffic Safety Administration)
  Good for: crash investigations, recalls, ADS-related defect investigations, safety reports, standing general orders on crash reporting.
  Search terms: “ADS crash report”, “standing general order”, “investigation”.

• NTSB (US National Transportation Safety Board)
  Good for: deep-dive accident investigations and recommendations.

• FMVSS / DOT / FHWA / USDOT
  Good for: policy, safety guidance, and transportation research in the US.

• UNECE WP.29 and related bodies
  Good for: international vehicle regulation and automated driving standards.

• UK DfT / Centre for Connected and Autonomous Vehicles (CCAV), DVSA, AAIB where applicable
  Good for: testing guidance, policy, and safety frameworks.

• Transport Canada, Australian road safety agencies, EU/European Commission
  Good for: broader international regulatory perspectives.

---

2. Independent transportation safety institutes

These often provide more analysis than agencies and are good for methodology and safety critique.

• IIHS (Insurance Institute for Highway Safety)
  Good for: independent evaluations, safety systems, driver assistance, crashworthiness, consumer-facing safety claims.

• TRB / National Academies / Transportation Research Board
  Good for: peer-reviewed conference proceedings, committee reports, and research synthesis.

• RAND Corporation
  Good for: policy analysis and scenario-based assessments of AV deployment.

• European Transport Safety Council (ETSC)
  Good for: policy and road safety analysis.

---

3. Academic publishers and conference proceedings

Best for technical limitations, validation methods, perception/planning safety, and system verification.

• IEEE Xplore
• ACM Digital Library
• Springer / Nature / Elsevier journals
• arXiv for preprints, with the caveat that it’s not peer reviewed

Useful search topics:

• “autonomous vehicle safety validation”
• “ADS scenario-based testing”
• “safety assurance case”
• “ODD limitations”
• “runtime monitoring”
• “formal verification autonomous driving”
• “sensor fusion failure modes”

Key venues:

• IEEE Intelligent Vehicles Symposium
• IEEE ITSC
• Safety-Critical Systems Symposium
• RSS, CoRL, ICRA, IROS for relevant robotics/autonomy safety work

---

4. Standards and certification organizations

Very useful for understanding how safety claims should be structured.

• ISO

  • ISO 26262: functional safety
  • ISO 21448 (SOTIF): safety of intended functionality
  • ISO/SAE 21434: cybersecurity for road vehicles

• SAE International

  • J3016: automation levels definitions
  • Other AV/ADAS guidance and taxonomy

• UL Standards & Engagement Good for: safety engineering and certification-oriented analysis

• ANSI / IEC / IEEE standards efforts where relevant

---

5. Industry safety reports and disengagement / incident data

Useful, but should be read critically because they may be selective or marketing-oriented.

• Waymo Safety Reports
• Cruise / GM public safety materials (where available)
• Tesla vehicle safety reports
  Note: these are not equivalent to independent AV safety assessments and often use company-defined metrics.

What to look for:

• definitions of miles driven
• what counts as a crash or disengagement
• operational design domain (ODD)
• comparison baseline
• whether the report includes human driver comparison data

---

6. Insurance and risk analysis publications

Useful for claims about real-world risk, claims frequency, and liability.

• Swiss Re
• Munich Re
• Lloyd’s
• Other insurance risk modeling groups

These can be good for systemic risk and deployment analysis, though not always deeply technical.

---

What kinds of sources are most credible for safety claims?

For evaluating a claim like “our system is safer than a human driver,” the strongest sources usually combine:

1. Independent crash statistics
2. Clear operational domain
3. Comparable baseline
4. Transparent methodology
5. Peer review or regulatory oversight
6. Evidence on limitations and edge cases

Red flags to watch for

Be cautious if a publisher or report:

• only reports “miles driven” without crash context
• uses vague “safety disengagement” definitions
• compares highway AV miles to city human driving without adjustment
• excludes low-speed or property-damage crashes
• omits the ODD or weather/road constraints
• relies solely on company-controlled data

---

If you want a practical reading list

A strong starting stack would be:

1. NHTSA + NTSB for incidents and regulations
2. IIHS for independent evaluation
3. ISO 26262 + ISO 21448 (SOTIF) + SAE J3016 for framework/definitions
4. IEEE / ACM papers for technical limitations and validation
5. Waymo safety reports and similar company reports for claims, but cross-check them critically
6. TRB / National Academies for synthesized research and policy

If you want, I can also give you:

• a ranked list by credibility,
• a list focused on US sources only, or
• a starter bibliography on AV safety validation and limitations.