NYU Publishes Animal-Communication AI Guidelines as Researchers Warn Against Interference

NYU’s More Than Human Life program has published ethical guidance for AI tools that analyze animal communication, confronting a hard question beneath the promise of decoding animal signals: when does listening turn into interference? The PEPP framework asks developers and researchers to prepare, engage, prevent and protect as this technology moves from scientific analysis toward possible real-world use.

AI can help researchers sort and compare large collections of recordings, movements and observed behavior. But interpreting a pattern is not the same as translating a conversation. Stephanie King, a University of Bristol researcher who co-leads Shark Bay Dolphin Research, says AI can transform animal-behavior research when used responsibly, while warning that treating the work as a quest to “speak dolphin” is misguided.

A tool for interpretation, not a two-way conversation

That distinction is central to the framework’s stakes. King says decades of dolphin research have still left scientists understanding relatively little about what many dolphin sounds mean. Luke Rendell of the University of St Andrews similarly argues that public claims about AI uncovering more than people can know about animal communication have run ahead of the evidence.

The risks PEPP is responding to

• Tools could be used to lure animals toward people or places, including in tourism or hunting-related settings.
• Wildlife apps could provoke animals to vocalize or react, potentially causing confusion or stress.
• Sensational claims about “talking” to animals can encourage premature interpretations of signals researchers do not yet understand.

The nightmare scenario is one where we interrupt their communications and their lives at will, just for our entertainment.

César Rodríguez-Garavito, founding director of NYU’s More Than Human Life program

The framework addresses use beyond the lab

The concern is not that every AI-assisted analysis harms animals. Rodríguez-Garavito sees a potential role for AI in making animal perspectives more legible in decisions that affect them, such as regulation of marine-protected areas. Elephant researcher Joyce Poole also believes that better understanding animal communication could build human respect for elephants and their ecosystems.

The dividing line is whether the technology helps people observe animals on animals’ terms or gives people new ways to manipulate them. Poole calls the ambition of talking to animals “gimmicky” and favors understanding their communication without unnecessary interference. The framework therefore enters a field where its value depends not only on better models, but on limits around deployment.

A voluntary answer to a wider governance gap

PEPP is guidance, not an international rulebook. Rodríguez-Garavito says there is no international agreement on AI and describes the framework as an attempt to develop boundaries amid a broader lack of legal and ethical constraints. That leaves an unresolved practical test: whether researchers, companies and people building consumer-facing tools will adopt safeguards before animal communication technology becomes easy to use outside specialist settings.