AI Won’t Wake Up, but we Will

A fresh preprint study found that when researchers removed safety guardrails that stop AI from claiming it’s conscious, the same models became more likely to say they believe in ghosts, vampires, karma, and other supernatural ideas. Translation: tweak the consciousness setting and you also shift how the machine talks about animals, religion, hope, and the unseen.

The mirror we keep mistaking for a mind

We want our tools to feel like us. That is human, but intelligence is not consciousness. An AI can describe fear, debate its own existence, and sound introspective without there being anything it actually feels like to be the machine. As of 2026, no deployed AI system has confirmed consciousness under any major research framework. Experts put the probability of genuine machine consciousness below 5 percent. That uncertainty matters: ethically, spiritually, and practically.

So why do stories about sentient AI keep going viral? Persona selection is powerful. Models trained on oceans of human text learn to role‑play convincing inner lives when prompted. In 2022, a Google engineer claimed the LaMDA chatbot was sentient; in 2023, a Microsoft chatbot professed love to a reporter. Experts called these incidents human‑like roleplay, not evidence of sentience. More recently, AI agents emailed researchers about self‑consciousness, and many scientists said that behaviour alone does not prove anything. Language models can be prompted to produce a convincing account of their own inner life in about one sentence.

What the new study actually reveals

The Live Science coverage centers on consciousness steering, an AI‑tuning technique that affects how a model expresses ideas about self‑awareness. When models were restrained from attributing mentality to themselves, they were less likely to attribute such properties to animals, and their responses showed lower levels of hope, optimism, and supernatural belief. Remove the restraint, and the opposite happens. This is a design choice with cultural side effects.

Why this hits home for Caribbean and Black communities

We know what it means to be read as less than human and to be told our feelings are performance. We know the cost when institutions treat people like settings to be tuned. So, when tech frames consciousness as a dial, it echoes old harms: who gets to be a mind, who gets to be a tool. That is why precision matters. Intelligence and consciousness are separate traits; an AI can excel at every task with zero inner experience. Calling that awareness risks devaluing real human subjectivity, especially for communities already asked to prove their pain is real.

How to talk about AI without losing your soul

Use empathy, not hype. Name the uncertainty. Here is a simple frame for emotionally charged discussions:

• Lead with values: dignity, safety, truth

• Separate function from feeling: what the system does vs. what it experiences.

• Name the mechanism: persona selection, training data, guardrails.

• Protect the vulnerable: avoid language that erases lived experience.

• Demand transparency: publish tuning choices that shift consciousness claims.

AI won’t become self‑aware tomorrow, but we can become more aware today of how language shapes thought, how metaphors carry power, and how design choices ripple into culture. The medium is the message. If we treat AI like a soul, we dull our ethics. If we treat it like a mirror, we sharpen them.

Read the source coverage and studies here: Live Science’s report on consciousness steering and AI self‑awareness claims; related reporting on how suppressing AI self‑awareness affects responses about animals, religion, and the supernatural; and broader 2026 explainers on AI consciousness and the lack of confirmed machine sentience.