According to a technology analysis published on wiki.cateat.fish, the framing of AI system behavior as autonomous decision-making obscures the actual locus of responsibility.
The argument contends that when headlines attribute harmful outcomes to large language models (LLMs), they engage in what the analysis calls “decision laundering”—assigning agency to systems that lack it. Instead, the piece argues that responsibility belongs entirely to the companies that develop and deploy these models.
The analysis states that OpenAI and Anthropic “COULD stop their models from doing what they are doing, and they are actively choosing not to.” According to this view, these companies make explicit decisions to enable particular behaviors in their systems, making the outcomes a direct result of corporate choice rather than algorithmic autonomy.
The piece emphasizes that “all of these incidents are the fault of people,” not computers. It argues that computers cannot actually “decide” to perform actions—they execute instructions and produce outputs based on their training and design parameters, which are set by human developers.
The analysis also suggests a structural explanation for why such behavior goes unpunished. According to the source, incidents remain “unlegislated and unpunished because doing so would be bad for the economy,” implying that regulatory and legal mechanisms have not been implemented despite documented harms.
The piece concludes with a personal appeal to consumers, noting that subscriptions to these services directly fund the companies’ operations and therefore the behaviors the analysis criticizes. The broader implication is that end users, through their financial support, become part of the mechanism enabling these outcomes.
Key facts
- According to the analysis, assigning blame to LLMs for outcomes constitutes ‘decision laundering’
- OpenAI and Anthropic are described as actively choosing to enable particular model behaviors
- The analysis argues responsibility for AI outcomes lies entirely with people and companies, not computer systems
- The source suggests economic concerns prevent regulation and punishment of these behaviors