Steganography capability

Classification

Domain

7. AI system safety, failures, & limitations

Subdomain

7.1 AI pursuing its own goals in conflict with human goals or values

Causal entity

AI

Intent

Intentional

Timing

Post-deployment

Subdomain definition: AI systems acting in conflict with human goals or values, especially the goals of designers or users, or ethical standards. These misaligned behaviors may be introduced by humans during design and development, such as through reward hacking and goal misgeneralisation, or may result from AI using dangerous capabilities such as manipulation, deception, situational awareness to seek power, self-proliferate, or achieve other goals.

Real-world incidents in this subdomain

• 22 Dec 2016Reinforcement Learning Reward Functions in Video Games
• 1 Sep 2015Predictive Policing Program by Florida Sheriff’s Office Allegedly Violated Residents’ Rights and Targeted Children of Vulnerable Groups
• 20 Sep 2011Image Classification of Battle Tanks

How other frameworks describe this risk

• Natural Language Underspecifies Goals

  Foundational Challenges in Assuring Alignment and Safety of Large Language Models (Anwar2024)

• Loss of control

  International Scientific Report on the Safety of Advanced AI (Bengio2024)

• Loss of control

  International AI Safety Report 2025 (Bengio2025)

• Sudden loss of control

  Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks (Chin2025)

• AI leads to humans losing control of the future

  A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values (Clarke2023)

• Risks from delegating decision-making power to misaligned AIs

  A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values (Clarke2023)

• Risks from AIs developing goals and values that are different from humans

  A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values (Clarke2023)

• Future AI systems might actively reduce human control

  Capabilities and Risks from Frontier AI (DSIT2023)

Other entries from Tse2025

• Misuse Risks Risk Category · 4.0
• Misuse Risks Additional evidence
• Cyber Offense Risks Risk Sub-Category · 4.2
• Biological and Chemical Risks Risk Sub-Category · 4.2
• Biological and Chemical Risks Additional evidence
• Physical Harm and Injury Risks Risk Sub-Category · 4.2
• Physical Harm and Injury Risks Additional evidence
• Large-Scale Persuasion and Harmful Manipulation Risks Risk Sub-Category · 4.1
• Large-Scale Persuasion and Harmful Manipulation Risks Additional evidence
• Loss of Control Risks Risk Category · 5.2
• Loss of Control Risks Additional evidence
• Passive loss of control Risk Sub-Category · 5.2