Concern 234 · Working draft · editorial review in progress

Long-horizon task drift

Long-horizon task drift occurs when an AI agent moves away from the user's intended task as it works through many steps.

Understanding the concern

An agent is software that uses AI to choose actions, use tools, and respond to results. In a long task, a mistaken assumption can affect later choices. Summaries, interruptions, and changing information can also make the original constraints less prominent.

This does not require a hidden motive. An agent can drift through ordinary errors or weak tracking of instructions. Intermediate checks should compare progress with the original request, not merely confirm that another step was completed. Final success needs to be judged against the user's actual goal.

An example

A travel assistant gradually optimizes for the cheapest trip while forgetting that the traveler needed to arrive before a family event.

Illustrative unless identified as a documented case.

Words in plain English

AI
Software that learns patterns from examples or uses rules to perform tasks such as recognizing images or generating text. It can make mistakes and does not understand every situation as a person would.
Agent
An AI system that can choose and carry out several steps toward a goal, using the tools and access people give it.
Drift
A change over time in data or real-world conditions that can make a system less reliable.

Evidence and its limits

The benchmarks establish difficulty in end-to-end tasks and evaluation limitations. The travel example illustrates drift rather than a measured prevalence claim.

Explore these sources in the library →
Broader frameworks and background

These resources provide context. They do not establish every claim in this concern.

A question to discuss

At which points does the agent recheck the original goal and constraints?

Keep track of different views, what needs checking, and any next step in your own discussion notes.

Open discussion notes →
Principles for the conversation

Before you finish · reflect together

Virtues and principles to consider

Virtues are qualities we practice in how we treat people and make choices. They are not scores for judging someone. Use these ideas to explore what a thoughtful response could look like.

Prudence

Prudence means using careful judgment before and during action. It balances worthwhile opportunities with the need to avoid harm that could have been anticipated.

Start with a limited trial, set stopping points, and prefer choices that can be undone when uncertainty is high. Review what happens rather than assuming the original plan remains sound.

Ask together: What is a careful next step, and when should we stop or change course?

Humility

Humility means recognizing the limits of our knowledge and being willing to learn from others. Expertise can be valuable without answering every question.

State where evidence is weak, invite criticism, and listen to people with different experience. Leave room to revise a decision when new information appears.

Ask together: What might we be missing, and who could help us see it?

Stewardship

Stewardship means caring for resources and systems that others depend on, including people who will use them later. It looks beyond the benefit of a quick launch.

Plan for maintenance, energy use, data protection, and eventual shutdown. Ask who will look after the tool when the original team moves on and how future users can avoid inheriting hidden costs.

Ask together: What are we responsible for caring for over time?

Editorial ratings

These are editorial judgments from the register, not measured probabilities or predictions about your situation.

Severity
Major · 4/5
Likelihood
Documented · 3/5
Time horizon
Near
Register risk score
12
About the register and scoring →
More ways to explore

Search for more writing

For more specific or recent material, open a search below or ask for a web summary. AI-generated summaries can contain mistakes; open the original articles to check them.

Explore a proposed use of AI →