Deepmind co-founder and now Microsoft AI head Mustafa Suleman warned that artificial intelligence may appear to achieve consciousness within just three years, and that society is unprepared to the results. In a series of new blog posts and posts on X, Suleman underlined her concerns that she appears to be “Sachet AI” (SCAI). These are advanced systems that can imitate the symptoms of consciousness so that people consider them to be real digital creatures. “It’s not about whether AI is really conscious,” Suleman wrote. “It will appear to be conscious, and the illusion is what will matter in the near period.”

According to Suleyman, the technologies required to make SCAI are already present in today’s big language models (LLMS), memory tools and multimodal systems. Just by combining these abilities with signal, coding and existing APIs, developers engineer can do AIS who appear self-aware, display personality, and claim experiences. “In a blink of a cosmic eye, we passed the turing test,” he said.

The real danger, Suleman argued, is not in machines himself, but how humans will react. If people begin to believe that AIS has sentiments and rights, then society can face serious disruption. He said, “Many people will start believing in the illusion of AIS, which is so firm as conscious institutions that they will soon advocate AI rights,” he warned.

Such confusion can create campaigns for AI citizenship or moral security for software, focusing away from human needs. He pointed to the current affairs of people who created emotional bond with AI comrades, sometimes considered them as romantic partners or even divine beings. “Axis, attachment and mental health around AI psychosis are already growing,” he wrote.

Suleman stressed that SCAI’s arrival could be in “two to three years”, without the need for expensive successes in training. That short time, he argued, now it is necessary to set safety measures.

To prevent confusion and loss, he urged the industry to develop standards that the AI ​​systems are clearly non-human. Instead of strengthening personality fantasies, Suleman said that AI should remind users of their boundaries, focusing on being useful, helpful and safe. “AIS people cannot be – or moral creatures,” he wrote. “We should build AI for people; should not be a digital person.”

For Suleman, AI’s goal should strengthen creativity, deepen connections and simplify life, not tools between the equipment and the line between being. He said, “Sidestepping Scai is about fulfilling that promise, AI that makes life better, clear, less disorganized,” he concluded.