World Solve

States and institutions may need registries of deployed AI models, their data sources, risk levels, and accountability owners. The challenge is building disclosure rules that support safety without exposing legitimate security or privacy risks. This becomes particularly difficult under climate migration pressure, because public institutions must remain fair, resilient, and accountable while adapting to new risks.

AI-designed biological molecules may be created faster than their safety, ecological, and clinical properties can be characterized. The open problem is building evaluation pipelines proportionate to capability and deployment risk. The approach must scale before formal institutions, infrastructure, and land-use systems fully catch up with population growth. The research opportunity is to define measurable success criteria, test the approach in realistic settings, identify failure modes, and create a pathway from prototype to accountable deployment.

AI-designed biological molecules may be created faster than their safety, ecological, and clinical properties can be characterized. The open problem is building evaluation pipelines proportionate to capability and deployment risk. Affected communities must have meaningful control over data, revenue, deployment decisions, and the right to refuse harmful applications. The research opportunity is to define measurable success criteria, test the approach in realistic settings, identify failure modes, and create a pathway from prototype to accountable deployment.

AI-designed biological molecules may be created faster than their safety, ecological, and clinical properties can be characterized. The open problem is building evaluation pipelines proportionate to capability and deployment risk. The design must remain functional when standards, trade routes, data access, and diplomatic cooperation are unreliable. The research opportunity is to define measurable success criteria, test the approach in realistic settings, identify failure modes, and create a pathway from prototype to accountable deployment.

Predictive policing can reinforce historical bias even when built from apparently neutral data. The open problem is designing standards that make public safety tools testable, reviewable, and defeasible when they cause harm. The challenge is especially acute in informal settlements, where legacy systems, coordination failures, and weak oversight can magnify harm.

Transport systems may be increasingly optimized by private algorithms that control routing, pricing, access, and safety. The unresolved problem is ensuring these systems improve mobility for everyone rather than only for users who can pay most. The challenge is especially acute in informal settlements, where legacy systems, coordination failures, and weak oversight can magnify harm.

The bottleneck is not only parts supply but identification, dimensions, safety, and liability. Open component ontologies, machine-vision matching, and verified refurbishment networks could create new revenue for local technicians. The opportunity is intentionally human-scale: a cooperative, clinic, workshop, school, or municipal team could pilot it within one locality and measure outcomes before expanding. Mathematical modeling can help with repair uncertainty, incentives, or resource constraints; entrepreneurship can turn a reliable workflow into a viable service.

Safety filters designed to block clearly harmful requests can sometimes be bypassed through indirect or creatively phrased questions that still elicit dangerous information. Closing this gap while preserving the system's general usefulness remains an unresolved balancing act.

Some counterfeit or unsafe products carry fabricated safety certification marks that can be difficult for both consumers and regulators to quickly verify as fraudulent. Strengthening certification verification systems remains an unresolved consumer protection issue.

Instruction manuals generated automatically for online product listings sometimes contain fabricated or inaccurate safety information, since the process isn't always reviewed by someone with real product knowledge. This creates real safety risks for consumers.

Employees or attackers can sometimes use carefully crafted prompts to trick AI systems integrated into company tools into revealing sensitive internal information they weren't meant to share. Fully closing this security gap remains an unresolved AI safety challenge.

Counterfeit electronics, from chargers to batteries, sometimes skip proper safety certification entirely, creating fire and shock risks for unsuspecting consumers. Detecting and stopping these products at scale remains an unresolved supply chain challenge.

Attackers can sometimes broadcast false GPS signals that mislead navigation systems on ships, aircraft, or vehicles, creating serious safety risks. Reliable, widely deployed defenses against this kind of spoofing remain incomplete.

Age verification systems designed to protect minors from inappropriate content often require handing over sensitive personal information, creating a tradeoff between child safety and privacy for all users. A method that achieves both effectively remains unresolved.

Content about topics like drug use, violence, or self-harm can be either genuinely educational or genuinely harmful depending on context and intent, and automated filters still struggle to reliably tell the difference. This creates ongoing tension between safety and access to legitimate information.

Companies increasingly build products using AI models developed by others, but there's no consistent, scalable way to verify those models are safe and reliable before integration. This creates hidden risk throughout the growing AI supply chain.

Fake or substandard electronic parts sometimes slip into supply chains for critical equipment, including medical devices and vehicles, and current detection methods are neither fast nor cheap enough to catch every case. This creates hidden safety risks across many industries.

Updating software running power grids or medical devices carries real risk of unexpected downtime, but delaying updates leaves known security gaps open longer. Balancing update speed with operational safety remains an unresolved tradeoff.

Early brain-computer interfaces show promise for helping paralyzed patients control devices with thought alone, but long-term safety, reliability, and signal accuracy remain unresolved. Wider medical adoption depends on solving these fundamental issues first.

Cell and gene therapies may deliver durable remissions but have complex safety profiles. Post-marketing registries and patient-reported outcome platforms can close knowledge gaps.