Risks & Limits

AI Risks and Limits

Hallucinations, bias, transparency, ethics, privacy, authenticity, jobs, and future claims.

How to use this section

AI risk is not limited to dramatic future scenarios. Present systems can invent details, expose information, amplify unfair patterns, create misleading media, or shift responsibility onto people who cannot meaningfully challenge the result. These pages focus on practical safeguards.

Fluency is not evidence

A polished explanation can contain an invented source, incorrect number, or false premise. Important claims require independent verification.

Average accuracy can hide harm

Error types and consequences should be examined for relevant groups and situations, not compressed into one overall score.

Transparency must be useful

A technical description is not enough. People need to know when AI matters, what information influenced an outcome, and how to correct or appeal it.

Accountability lasts beyond launch

Data, users, threats, products, and operating conditions change. Named owners need monitoring, incident response, rollback, and retirement plans.

Guide library

AI Risks and Limits articles

Risks & Limits

AI Hallucinations and Verification

Learn why AI systems can invent facts, quotations, citations, or reasoning, and use a practical verification method before relying on generated content.

Risks & Limits

AI Bias and Fairness

Understand how data, labels, design choices, and deployment conditions can produce unfair AI outcomes, and why fairness requires more than removing sensitive fields.

Risks & Limits

AI Explainability and Transparency

Learn the difference between explaining a model, documenting a system, and giving people meaningful information about an AI-assisted decision.

Risks & Limits

Responsible AI and Ethics

A practical framework for purpose, proportionality, human oversight, safety, fairness, accountability, and stopping an AI system that causes harm.

Risks & Limits

AI Privacy, Security, and Copyright

Understand the data, security, confidentiality, consent, and copyright questions that arise when people train, connect, or use AI systems.

Risks & Limits

AI Detectors and Content Authenticity

Learn why AI-text detectors are uncertain, how image provenance can help, and why authenticity decisions should use evidence rather than a single score.

Risks & Limits

AI, Jobs, and Social Impact

Explore how AI changes tasks, job design, skills, access, concentration of power, and the distribution of benefits and harms.

Risks & Limits

The Future of AI: What We Know and What We Do Not

A grounded way to evaluate claims about future AI capability, adoption, regulation, labour, and social change without treating forecasts as facts.