IBM’s Watson supercomputer gave unsafe recommendations for treating cancer patients, according to documents reviewed by Stat. The report is the latest sign that Watson, once hyped as the future of cancer research, has fallen far short of expectations.

In 2012, doctors at Memorial Sloan Kettering Cancer Center partnered with IBM to train Watson to diagnose and treat patients. But according to IBM documents dated from last summer, the supercomputer has frequently given bad advice, like when it suggested a cancer patient with severe bleeding be given a drug that could cause the bleeding to worsen. (A spokesperson for Memorial Sloan Kettering said this suggestion was hypothetical and not inflicted on a real patient.)

“This product is a piece of s—,” one doctor at Jupiter Hospital in Florida told IBM executives, according to the documents. “We bought it for marketing and with hopes that you would achieve the vision. We can’t use it for most cases.”

The documents come from a presentation given by Andrew Norden, IBM Watson’s former deputy health chief, right before he left the company. In addition to showcasing customer dissatisfaction, they reveal problems with methods, too. Watson for Oncology was supposed to synthesize enormous amounts of data and come up with novel insights. But it turns out most of the data fed to it is hypothetical and not real patient data. That means the suggestions Watson made were simply based off the treatment preferences of the few doctors providing the data, not actual insights it gained from analyzing real cases.

According to IBM spokesperson Edward Barbani, Watson for Oncology started off by using real patient data. But this made it difficult to do updates each time the guidelines changed, so the scientists switched to using hypothetical cases. “Synthetic cases allow you to treat and train Watson on a variety of patient variables and conditions that might not be present in random patient samples, but are important to treatment recommendations,” he said.

Update July 31, 2018 1:30 p.m.: This article has been updated to include information from IBM.