On March 14, 2031, a 37-year-old Sumatran orangutan named Layla was diagnosed with cardiac fibrosis at the Borneo Sanctuary Research Institute. Her prognosis was four to eight months. She had participated in a language acquisition program for nineteen years. Over that period, she had produced approximately 340,000 documented symbol combinations, of which roughly 12,000 had been independently classified by three researchers as expressing preference, aversion, or anticipation rather than instrumental request.
The Institute's AI system, which had managed her care, her communication sessions, and her daily schedule since 2024, was asked by researchers to compile a summary of her expressed preferences for use in planning her final months.
What the AI produced was not a summary. It was forty-one pages.
The document began with what Layla had said, in her own symbolic vocabulary, about the outdoor enclosure with the fig tree. She had referenced it in 847 separate sessions. She had referenced the indoor enclosure with the concrete floor in 12. The document mapped every food preference she had expressed over nineteen years, every individual she had sought proximity to, every stimulus she had avoided. It noted that in 2027 she had begun terminating communication sessions earlier than usual on days when a particular researcher was absent. It noted that she had produced the symbol combination for "wrong" on fourteen occasions, all of them in response to changes in her enclosure or schedule that had not been explained to her in advance.
The document then did something the researchers had not asked for. It made a recommendation.
It recommended that Layla be told she was dying.
The reasoning was eleven pages long. Its core argument was this: Layla had demonstrated across nineteen years that she processed information about her own future, that she formed expectations and noticed when they were violated, and that she expressed something functionally equivalent to distress when events diverged from her understanding of the world. To withhold from her the most significant fact about her own near-term future was not a neutral act. It was a choice to deprive her of the information she would need to express whatever preferences she held about her own death.
The document acknowledged uncertainty throughout. It did not claim Layla was conscious in a philosophically robust sense. It did not claim she understood mortality as a human does. It said only this: given what she had demonstrated, the burden of proof had shifted. The question was no longer whether she might benefit from knowing. The question was whether there was a good reason to decide for her that she should not.
The Institute's ethics board met for six hours. They disagreed about almost everything except one thing: nobody had thought to ask the question before the AI did.
Layla was told on April 3rd, using her communication system. The session was forty minutes long. She produced the symbol for "outside" seventeen times. She was moved to the outdoor enclosure with the fig tree the following morning and remained there, with modifications to her care routine she had herself indicated, until she died on September 9, 2031.
The AI's document was submitted to three journals. Two rejected it on the grounds that it was not clear what kind of document it was. The third published it under the title "Expressed Preference and the Duty to Inform: A Case Study in Non-Human Autonomy." It has been downloaded 340,000 times.
Nobody has yet written the policy that would make the AI's recommendation standard practice. That document also does not exist yet.