There are many approaches to developing a taxonomy of AI risks. One framing divides risks by misuse, accident, and structural. It can be helpful to consider cause (e.g. accident or misuse), harm (e.g. economic, political, psychological, something else?), and mechanism (e.g. improved surveillance empowering autocracies).
I’ve adopted MIT’s seven-domain taxonomy for the following running list of risks associated with AI systems:
1. Discrimination & Toxicity
- inequality
- misrepresentation
- exposure to toxic content
- embedding of bias
2. Privacy & Security
- erosion of privacy
- cyber threat
3. Misinformation
- pollution of information ecosystem and loss of consensus reality
- mass manipulation
4. Malicious actors & Misuse
- disinformation
- fraud and targeted manipulation
- cyber attacks
- weapon development or use (including autonomous weapons and chemical, biological, radiological, nuclear, and explosive weapons)
5. Human-Computer Interaction
- over-reliance and dependence
- cognitive degradation
- disempowerment and loss of agency
- further fragmentation of public discourse
- erosion of peer review and scholarly quality control
6. Socioeconomic & Environmental Harms
- centralisation of power and unfair distribution of benefits
- empowered autocrats
- labour market shocks
- governance failures
- empowered autocrats
- increased inequality
- institutional overwhelm
- geopolitical pressures
7. AI system safety, failures, & limitations
- dangerous emergent capabilities
- loss of control
- critical infrastructure vulnerabilities
- multi-agent risks
- autonomous AI systems pursuing goals in conflict with human safety
see also: MIT AI Risk Initiative