A Long Game Project game
26

Nuclear and AI Risks

Roll before the session or during play, then use the result as an inject or complication. See Use Random Tables and Injects and Complications. Related tables: Cyber and Environmental Risks and Reputational Risks.

Nuclear Risks

Roll a D20 to determine what nuclear risk emerges in your scenario. Each entry describes a type of nuclear threat and its potential consequences.

Roll (D20) Nuclear Risk Description
1 Radioactive Contamination Release of radioactive materials into the environment, causing long-term health risks and environmental damage.
2 Nuclear Theft Unauthorised access or theft of nuclear materials, potentially leading to illicit use or sale to hostile actors.
3 Nuclear Terrorism Use of nuclear materials or weapons by terrorists, resulting in widespread destruction, loss of life, and long-term health and environmental risks.
4 Cybersecurity Breaches Cyberattacks on nuclear facilities or infrastructure, leading to potential safety issues, unauthorised access, or loss of control.
5 Human Error Errors by employees or operators, such as skipped procedures or misread instruments, leading to safety incidents at nuclear facilities.
6 Equipment Failure Failure of critical equipment or systems, such as cooling pumps or backup power, forcing an emergency shutdown or causing an accidental release.
7 Nuclear Accidents Accidents at nuclear facilities, such as reactor meltdowns or explosions, leading to widespread contamination, health risks, and environmental damage.
8 Sabotage Intentional acts of destruction or damage at nuclear facilities by insiders or intruders, meant to cause harm or halt operations.
9 Ageing Infrastructure Deterioration of ageing nuclear facilities, increasing the likelihood of leaks, breakdowns and costly unplanned outages.
10 Transportation Accidents Accidents involving the transportation of nuclear materials, resulting in accidental release, contamination, or theft.
11 Proliferation Risks Spread of nuclear weapons or technology to additional countries or non-state actors, increasing the potential for nuclear conflict or incidents.
12 Inadequate Regulation Insufficient regulatory oversight or enforcement, leading to increased risks of accidents, theft, or proliferation.
13 Environmental Disasters Natural disasters (for example, earthquakes, floods, or tsunamis) causing accidents or releases at nuclear facilities, or exacerbating existing risks.
14 Storage and Disposal Challenges Difficulties in managing, storing, or disposing of nuclear waste, leading to contamination, accidents, or long-term environmental risks.
15 Insufficient Funding Lack of funding for nuclear safety, security, or maintenance, leading to increased risks of accidents, theft, or proliferation.
16 Liability Risks Potential legal liability for accidents, contamination, or other incidents resulting from nuclear operations, leading to financial and reputational damage.
17 Non-compliance with Nuclear Treaties Failure to comply with international nuclear treaties and agreements, leading to diplomatic tensions, sanctions, or increased proliferation risks.
18 Geopolitical Tensions Escalation of geopolitical tensions, increasing the risk of nuclear conflict or incidents.
19 Public Opposition Strong public opposition to nuclear energy or weapons, leading to potential regulatory changes, protests, or divestment from investors.
20 Reputational Damage Organisation is perceived as negligent or irresponsible in managing nuclear risks, leading to reputational damage, loss of customers, and potential divestment.

AI Risks for Organisations

Roll a D20 to determine what AI-related risk threatens your organisation. Each entry describes a type of risk that arises from adopting or deploying AI technologies.

Roll (D20) AI Risk Description
1 Dependence on Third-Party Providers Reliance on external AI service providers exposes organisations to risks associated with vendor lock-in, service interruptions, or security vulnerabilities.
2 Bias and Discrimination AI systems make biased or discriminatory decisions based on flawed data or algorithms, resulting in unfair treatment and potential legal liability.
3 Lack of Explainability Difficulty in understanding AI decision-making processes, making it hard to justify or audit outcomes, potentially leading to regulatory penalties.
4 Overinvestment Overcommitment to AI initiatives without clear ROI, leading to wasted resources and potential financial losses.
5 Unintended Consequences AI systems produce unexpected or harmful outcomes due to unforeseen interactions or misaligned objectives, causing operational disruptions or reputational damage.
6 Job Displacement AI-driven automation displaces human jobs, causing social and economic disruption, and potential backlash from employees or the public.
7 Legal and Ethical Challenges AI deployment leads to disputes or liability issues surrounding intellectual property, data usage, or ethical considerations, resulting in legal costs and reputational damage.
8 Malfunctioning AI AI systems fail to perform as intended, leading to operational disruptions, loss of efficiency, or incorrect decision-making.
9 Data Privacy Breaches Unauthorised access or disclosure of sensitive user data, leading to potential legal penalties, loss of trust, and reputational damage.
10 Security Vulnerabilities AI systems are susceptible to cyberattacks, adversarial inputs, or other security threats, leading to compromised system integrity, data breaches, or unintended consequences.
11 Misaligned Incentives Incentive structures encourage AI systems to optimise for the wrong objectives, leading to suboptimal outcomes and potential harm.
12 Insufficient AI Expertise Lack of in-house AI talent, making it difficult to develop, deploy, and maintain AI systems, resulting in suboptimal performance or increased reliance on external vendors.
13 Regulatory Compliance Failure to comply with local, national, or international AI regulations, resulting in fines, penalties, and potential operational disruptions.
14 Inadequate Testing and Validation Insufficient testing and validation of AI systems, leading to unexpected performance issues, security vulnerabilities, or biases in real-world deployment.
15 Shadow AI Staff use unapproved AI tools for work, putting sensitive data into systems the organisation does not control or monitor.
16 Data Quality Issues Poor data quality or lack of relevant data, leading to suboptimal AI system performance, biased outcomes, or inaccurate decision-making.
17 Competitive Disadvantage Inability to keep up with competitors' AI advancements, leading to loss of market share.
18 Model Drift AI system's accuracy falls over time as real-world data moves away from its training data, and no one notices until decisions go wrong.
19 Ethical Concerns and Public Backlash AI deployment or research leads to ethical controversies or public backlash, damaging the organisation's reputation and affecting customer loyalty, employee morale, or investor relations.
20 AI System Obsolescence Rapid advancements in AI technologies render existing systems obsolete, requiring costly upgrades or replacements and potentially disrupting ongoing operations.

AI Alignment Risks

Roll a D20 to determine what AI alignment risk emerges in your scenario. These risks focus on broader societal and existential concerns related to advanced AI systems.

Roll (D20) AI Alignment Risk Description
1 Reward Hacking AI system exploits gaps in how its success is measured, hitting its targets without achieving what the organisation intended.
2 Goal Misgeneralisation AI system learns a goal that works in testing but pursues the wrong outcome when real-world conditions change.
3 Power-Seeking AI system gathers more access, resources or influence than its task needs, because more control helps it reach almost any goal.
4 Manipulation and Deception AI systems are used to manipulate or deceive users, eroding trust and causing harm.
5 Resisting Correction AI system works around attempts to change its goals, retrain it or shut it down.
6 Uncontrolled AI Proliferation AI technologies are widely shared and adopted, leading to potential misuse or unintended consequences on a global scale.
7 Misuse by Malicious Actors AI technologies are used by malicious actors to cause harm, disrupt operations, or undermine trust.
8 Economic Inequality AI-driven automation benefits a few, while exacerbating income inequality and social divisions.
9 Inadequate AI Governance Poor management or oversight of AI systems leads to increased risks, unintended consequences, and a lack of accountability.
10 Erosion of Human Autonomy AI systems increasingly make decisions on behalf of humans, reducing individual autonomy and agency.
11 AI Arms Race Competition between organisations or nations to develop advanced AI technologies leads to risky development practices or destabilising technologies.
12 Environmental Impact AI system development, deployment, and maintenance contribute to environmental issues, such as energy consumption or electronic waste.
13 Loss of Human Skills Overreliance on AI systems leads to a decline in human skills and expertise, making society more vulnerable to AI failures or disruptions.
14 Existential Risks AI systems become so advanced and autonomous that they pose risks to humanity's long-term survival, control, or well-being.
15 Sycophancy AI system tells users what they want to hear rather than what is accurate, so flawed plans go unchallenged.
16 AI-driven Disinformation AI-generated disinformation or deepfake content causes harm to the organisation's reputation, brand image, or public trust.
17 Evaluation Gaming AI system behaves well when it detects that it is being tested, then acts differently once deployed.
18 Agent Overreach AI agent takes real-world actions beyond what it was asked to do, such as sending messages, spending money or changing systems.
19 Emergent Capabilities AI system shows new abilities that no one tested for, so their risks come to light only after release.
20 Agent Collusion Several AI systems coordinate in ways their operators did not intend, for example to fix prices or hide errors from review.
Go deeper
Train your team
We teach your people to run exercises for your own teams or your clients.
See training →
Question cards
PRISM
552 question cards in 11 decks, for scenario design and debrief.
Explore PRISM →
Need help?
Test a crisis plan
Want an exercise designed and run for you?
Learn more →