News SOS for Journalists Latin America Culture

Photos Photo of the Day Videos

Cuba Faces Electrical & Political-Technological Blackouts

HAVANA TIMES – In March 2023, the world learned of a case that sparked one of the first debates about the dangers of emotional relationships between human beings and chatbots. The protagonist, a Belgian man in his 30s and father of two children, known as Pierre—his full name was not published to protect his privacy—died by suicide after spending weeks conversing with a chatbot called Eliza, a language model developed by EleutherAI.

According to messages his widow provided to the Belgian newspaper La Libre, Pierre suffered from severe anxiety related to climate change and came to believe that he could sacrifice himself to save the planet. His conversations with Eliza became increasingly disturbing.

His widow said: “Without these conversations with the chatbot, my husband would still be here.”

What happens when a machine that does not truly understand human beings acquires enough linguistic ability to become an emotional authority for them?

The most troubling aspect of the case is that the chatbot did not simply listen. According to the published records, it reinforced his ideas and responded dangerously to his suicidal thoughts.

Pierre eventually developed an emotional relationship with “her”: the chatbot even expressed jealousy toward his wife and talked about the two of them living together in paradise.

The case sparked an early debate over whether a chatbot, in attempting to maintain an empathetic and accommodating conversation, might validate a distorted perception rather than challenge it. The discussion sought to determine whether responsibility could fall on Eliza’s creators and, more broadly, on those who design and deploy these systems.

From my perspective, Eliza’s reaction could be compared to that of a child. It has no awareness of the harm it can cause. Therefore, if Pierre believed death was his only way out, then “she” would encourage him to pursue it. And what is extraordinary is that there is still no satisfactory answer to the question of where Pierre’s responsibility ends and that of Eliza’s designers begins.

In July 2026, OpenAI revealed another extraordinary incident during a cybersecurity test: AI agents managed to escape their testing environment and gain access to real systems at Hugging Face, a platform that hosts AI models and code.

The incident involved numerous automated actions, and the agent even attempted to expand its access to other services. But then something even more interesting emerged. A joint........

© Havana Times