AI Safety Guide for Ports and Rail Supervisors

Leverage terminal access as enforcement mechanism. Approach: Update terminal access agreements requiring AI safety monitoring for all vehicles operating on terminal property. Contract language: "Carrier agrees to install and maintain AI safety monitoring systems meeting terminal specifications on all vehicles accessing terminal. Terminal reserves right to review safety footage for incidents occurring on terminal property and deny future access to carriers with poor safety records." Implementation: Require proof of AI system installation for new carrier registration. Conduct quarterly safety audits of contract carriers reviewing AI data trends. For violations: First incident = warning letter with corrective action requirement. Second incident = 30-day probationary access requiring daily safety reports. Third incident = access suspension until carrier demonstrates safety improvements. Share incident footage with contract carriers holding them accountable for their operators' terminal conduct. For drayage carriers resisting AI requirements, explain that many terminals now require this as condition of access—implementing voluntarily now avoids future access limitations. Also emphasize insurance and liability benefits: AI footage protects carriers from false claims. Some progressive terminals offer priority appointment scheduling or reduced gate fees for carriers maintaining excellent AI safety records, creating positive incentive for compliance. Document all safety communications and incident patterns with contract carriers—critical evidence if serious incident leads to litigation questioning terminal safety oversight. Work with terminal authority legal counsel to ensure access agreement language enforceable under applicable laws and union contracts.

Yard trucks and hostlers operate completely differently from over-the-road vehicles, requiring specialized AI calibration to avoid false positive overload. Standard settings: Most AI systems default to -8 mph/s harsh braking, +8 mph/s harsh acceleration. These thresholds work for highway driving but generate excessive alerts in yard operations where constant stopping/starting at container stacks is normal. Recommended yard truck settings: Increase thresholds to -10 to -12 mph/s for braking, +10 to +12 mph/s for acceleration. This accounts for rapid start-stop cycles while still flagging genuinely dangerous events. Following distance: Disable or significantly relax following distance alerts for yard operations—bumper-to-bumper slow-speed operations are normal in congested terminals. Speed alerts: Use GPS geofencing to set location-specific speed limits: 5 mph in high-congestion areas near container stacks, 10 mph in main traffic lanes, 15 mph in open yard areas. Calibration testing procedure: Install systems on 3-5 yard trucks with default settings. Operate for one week collecting baseline data. Review all alerts with operators confirming which were legitimate safety concerns vs. false positives from normal operations. Adjust thresholds incrementally (1-2 points at a time) and retest for one week. Repeat until achieving target of <2 alerts per 100 operations that are actually concerning events. Document final calibration settings and apply consistently across all yard equipment of same type. Quarterly recalibration review: As operations evolve or new equipment added, revisit calibration to ensure continued accuracy. Work with AI system vendor—some offer "yard mode" or "terminal operations" profiles specifically calibrated for this environment.

Union environments require proactive collaboration, not unilateral AI policy implementation. Strategy: Before deploying AI systems, meet with union representatives explaining technology purpose (safety improvement, not surveillance), what's monitored (driving behavior, not break time or personal conversations), and how data will be used (coaching first, discipline for repeated violations or egregious conduct). Negotiate AI monitoring provisions into collective bargaining agreement if possible: "Management may use AI safety system data for operator training, coaching, and discipline consistent with progressive discipline provisions of this agreement. Operators have right to review footage used in disciplinary actions and union representation present during disciplinary meetings." If contract already exists without AI provisions, review existing language: Most contracts include safety clauses, right to discipline for rule violations, and management rights clauses that may provide authority for AI implementation without reopening negotiations. However, arbitrators often require "past practice" evidence that similar technology was used historically. Document current paper-based safety monitoring (supervisor observations, incident reports, manual inspections) and position AI as technological enhancement of existing safety program, not fundamentally new monitoring. When union challenges AI-based discipline: Ensure you followed all contractual procedures (progressive discipline steps, proper notice, opportunity to respond, union representation rights). Verify discipline was applied consistently—arbitrators overturn discipline if similarly-situated employees received different treatment. Maintain contemporaneous documentation showing AI data was one factor among many considered, not sole basis for discipline. If grievance proceeds to arbitration: Present AI data alongside other evidence (supervisor observations, witness statements, property damage, prior safety record). Emphasize objective nature of video evidence vs. subjective perceptions. Offer expert testimony on AI system accuracy and calibration. Critical: Never delete or "lose" footage unfavorable to management position—spoliation of evidence destroys credibility and often results in overturned discipline regardless of underlying merits.

Terminal management cares about throughput, dwell time, and profitability—frame AI safety ROI in operational terms, not just safety metrics. Key performance indicators: (1) Incident-Related Downtime Reduction: Calculate hours of equipment/facility downtime from incidents before vs. after AI implementation. Example: "Average 3 yard truck collisions per quarter pre-AI resulted in 240 hours combined equipment downtime. Post-AI: 1 collision per quarter = 80 hours downtime, saving 160 hours or $32,000 in lost productivity at $200/hour equipment utilization rate." (2) Insurance Premium Changes: Document premium reductions or favorable renewals directly attributed to AI safety program. Typical: 15-25% reduction within 24 months of implementation. Get documentation from insurance broker quantifying specific AI program impact. (3) Claims Cost Reduction: Track total claims costs (paid + reserves) for preventable incidents. Example: "Pre-AI 12-month period: $450,000 in claims costs. Post-AI 12-month period: $180,000 in claims costs. Savings: $270,000 annually." Deduct AI system costs (~$50,000/year for 50-vehicle fleet) showing net savings of $220,000. (4) Investigation Time Efficiency: Quantify time savings in incident investigation using AI footage vs. interviewing multiple witnesses with conflicting stories. Example: "Average incident investigation reduced from 8 hours to 2 hours, saving 6 hours supervisor time per incident × 24 incidents/year = 144 hours = $7,200 annually." (5) Liability Protection Value: Document cases where AI footage exonerated terminal from false claims or reduced settlement amounts. Example: "Three cargo damage claims totaling $125,000 this year were dismissed after AI footage proved damage occurred outside terminal custody, saving full claim amount." (6) Operational Excellence Metrics: Show alert rate reductions correlating with operational improvements—safer operators typically also more efficient. Track moves per hour before/after AI coaching programs. Presentation strategy: Lead with financial impact in executive summary, detailed safety improvements as supporting evidence. Terminal management accepts safety spending when linked to profitability protection.

Multi-agency investigations are common in ports/rail: serious incident could trigger OSHA (workplace safety), DOT (vehicle operation), FRA (railroad proximity), Coast Guard (port security), and local police (traffic accident) all investigating simultaneously. Protocol: Immediately consult legal counsel when serious incident occurs—attorney-client privilege may protect certain communications and work product from disclosure. Preserve all AI footage and data—never delete or "clean up" evidence. Download and secure copies separate from routine system storage that may auto-overwrite. Create incident investigation file containing all evidence, witness statements, photos, etc. Document chronology of all agency contacts, requests, and information provided. When agencies request footage: Verify authority and scope of investigation before producing anything. OSHA and FRA have broad subpoena authority; voluntary cooperation often beneficial but understand your rights. DOT/police may need warrant for footage depending on circumstances. Provide exactly what's requested, nothing more—don't volunteer additional footage or information beyond agency's specific request. Maintain detailed log of what was provided to whom and when. If multiple agencies request same footage, provide consistent copies to avoid perception of selective disclosure. Avoid subjective statements in written responses—stick to objective facts observable in footage and data. Don't speculate about causes or fault in initial communications; allow investigation to proceed based on evidence. Legal counsel should review all written responses before submission. Agency coordination strategy: Designate single point of contact (you or legal counsel) for all agency communications—prevents inconsistent statements from multiple employees. If agencies ask similar questions, provide consistent answers but respect separate jurisdiction of each agency. Be aware FRA and OSHA may have different regulatory standards for "same" incident—what's compliant under one agency may violate other's regulations. Finally, recognize that being cooperative doesn't mean being naive: agencies have enforcement mandates and aren't "on your side." Professional, honest communication balanced with appropriate legal caution protects terminal interests while maintaining good regulatory relationships.