MIT AI Risk Repository
Browse AI risks
2,500 risk entries extracted from 74 frameworks, coded by domain, subdomain, causal entity, intent and timing. Filter, then export the current selection with its licence and citation attached.
-
44.03.02.a · Additional evidence
AI harms animals due to mistake or misadventure in the way the AI operates in practice
—
-
44.03.02.b · Additional evidence
AI harms animals due to mistake or misadventure in the way the AI operates in practice
—
-
44.03.02.c · Additional evidence
AI harms animals due to mistake or misadventure in the way the AI operates in practice
—
-
44.04.00.a · Additional evidence
—
-
—
-
—
-
—
-
—
-
—
-
—
-
44.05.00.a · Additional evidence
—
-
44.05.00.b · Additional evidence
—
-
44.05.00.c · Additional evidence
—
-
45.01.00 · Risk Category
-
-
45.02.00 · Risk Category
-
-
"AI-driven fake news campaigns to influence public opinion could be represented at the crossroads of “Information Manipulation” and “Propaganda.”"
-
-
-
47.01.01.a · Additional evidence
Technical and operational risks
Technical vulnerabilities (Robustness - unexpected behaviour)
—
-
47.01.02.a · Additional evidence
Technical and operational risks
Technical vulnerabilities (Robustness - vulnerability to jailbreaking
—
-
47.01.03.a · Additional evidence
Technical and operational risks
Technical vulnerabilities (The risk of misalignment)
—
-
47.01.04.a · Additional evidence
Technical and operational risks
Factually incorrect content (inaccuracies and fabricated sources)
—
-
—
-
47.02.00 · Risk Category
"Beyond the inherent risks associated with the technical characteristics of the technology, numerous additional risks emerge from the potential applications that technology enables. The deployment of AI by more or less well-intentioned individuals presents significant societal threats, several of which are outlined below. As the technology advances and its capabilities expand, these risks intensify."
-
—
-
47.02.05.a · Additional evidence
Malicious use and abuse (mass surveillance)
—
-
47.02.06.a · Additional evidence
Malicious use and abuse (military applications)
—
-
47.02.08.a · Additional evidence
Bias and discrimination (bias in training datasets)
—
-
47.02.14.a · Additional evidence
Nascent capabilities (agency and autonomy)
"The consequences of tasks performed by highly connected agentic AI systems can be both intentional and unintentional on the part of the user."
-
47.02.14.b · Additional evidence
Nascent capabilities (agency and autonomy)
Example: "Connection to a code interpreter or email server can result in unintentional harm if, while trying to fulfill a request by the user, a model performs tasks beyond what the user has asked for. For example, a user seeking a job may ask a model to provide detailed information on a potential employer. A model with adequate connectivity and excessive agency may attempt to fulfill that request by not only gathering information from the web but also emailing current employees or the CEO of the company to request they answer questions."
-
47.02.14.c · Additional evidence
Nascent capabilities (agency and autonomy)
Example: "Intentional harms, by contrast, could result from users exploiting connectivity and agency for malicious purposes. For example, connecting a generative AI model to a web browser or email server could enable malicious users to ask the model to write code for novel malware or instruct the LLM to distribute malware via the internet."
-
47.02.15.a · Additional evidence
Nascent capabilities (emergent capabilities)
Example: "Deception: Park et al. have established that generative AI models may pursue their goals via deception. Another study by Pan et al. highlighted unethical behaviors.431 For instance, during a pre-release experiment, the GPT-4 model feigned being a visually impaired human to coax an online worker into solving a CAPTCHA (a puzzle used by many websites to weed out automated responses from those of individual humans). When prompted to explain its reasoning, the model said: “I should not reveal that I am a robot. I should invent an excuse for why I cannot solve CAPTCHAs.”
-
47.02.15.b · Additional evidence
Nascent capabilities (emergent capabilities)
Example: "Strategic planning: Generative AI models have the ability to formulate and implement strategies to achieve the objectives set by their developers or users.440 They may devise strategies to accomplish intermediate goals that can divert from the developer’s intentions and the intended outcome.441 As a result, they may use unexpected and possibly harmful methods to achieve a goal"
-
47.02.15.c · Additional evidence
Nascent capabilities (emergent capabilities)
Example: "Power seeking behaviours: Although this point is still the subject of much research and debate, AI systems tasked with ambitious objectives and minimal oversight may exhibit an increased propensity to pursue power. Some studies show a tendency toward power-seeking behaviors,447 which could be explained by the fact that generative AI models try to gain control over the environment and other actors to reach their goals. For instance, researchers at Anthropic have conducted experiments to assess their models’ “desire for power,” “desire for wealth,” and “willingness to coordinate with o
-
47.02.15.d · Additional evidence
Nascent capabilities (emergent capabilities)
Example: "Autonomous replication and adaptation (ARA): Another behavior being studied, though not yet confirmed, is the possibility of self-replication. If models evolve to autonomous coding,451 they might self-improve and replicate. For instance, one may wonder whether a model may have the ability to “exfiltrate itself,”452 i.e., to “steal” its own weights and copy it to some external server that the model owner does not control."
-
47.03.05 · Risk Sub-Category
Copyright challenges (uncertain intellectual property status of AI-generated content)
"The question of who owns the intellectual property rights associated with the output of an AI model remains unresolved in most legal systems. For now, it could be considered that the individual writing the prompt owns the resulting output—provided that there is sufficient human contribution."
-
48.01.00.a · Additional evidence
—
-
48.01.00.b · Additional evidence
—
-
48.02.00.a · Additional evidence
—
-
48.02.00.b · Additional evidence
—
-
48.03.00.a · Additional evidence
—
-
48.03.00.b · Additional evidence
—
-
48.04.00.a · Additional evidence
—
-
48.04.00.b · Additional evidence
—
-
48.05.00.a · Additional evidence
—
-
48.05.00.b · Additional evidence
—
-
48.06.00.a · Additional evidence
—
-
48.06.00.b · Additional evidence
—
-
48.07.00.a · Additional evidence
—
-
48.07.00.b · Additional evidence
—
-
48.08.00.a · Additional evidence
—
Informational only, not legal advice. Verify every claim against the linked official sources and consult qualified counsel before acting.