There is an association between sleep and dementia, with long sleep duration providing a clearer signal than short sleep duration. Reverse causality is a major caveat: disturbed sleep may be an early sign of dementia rather than a cause. Genetic research finds little evidence for a causal relationship.
The link between sleep and dementia exists, but it is considerably more nuanced than newspaper headlines often suggest. The most striking signal from the research is not short sleep but rather long sleep: people who consistently sleep a great deal appear to have a clearly higher risk of dementia. Short sleep showed no statistically demonstrable effect in the analyses, so the fear that a few hours less sleep damages your brain is less well supported than is often implied.
The biggest caveat is reverse causality. Disrupted or prolonged sleep may be an early sign of a process in the brain that is later recognised as dementia, rather than a cause of it. Research that tests this question more directly by using genetic variants as a starting point finds little evidence that sleep problems actually cause dementia. The association is therefore real, but whether it reflects a cause-and-effect relationship or an early symptom has not yet been clarified.
What does this mean for your concerns? If someone you know suddenly starts sleeping much more, sleeps restlessly, or has confused episodes at night at an older age, that is something to discuss with a doctor -- not because sleep is dangerous in itself, but because it can be one of the early signals that a GP will want to look at in combination with other things. People who simply sleep somewhat less than the textbook recommends have no reason to worry based on this evidence.
Moderate evidence, 5 source(s); the direction is probable but not firmly proven.