This review discusses three different associations between cannabinoids and cancer. First, it assesses evidence that smoking of cannabis preparations may cause cancers of the aerodigestive and respiratory system. There have been case reports of upper-respiratory-tract cancers in young adults who smoke cannabis, but evidence from a few epidemiological cohort studies and case-control studies is inconsistent. Second, there is mixed evidence on the effects of THC and other cannabinoids on cancers: in some in vitro and in vivo studies THC and some synthetic cannabinoids have had antineoplastic effects, but in other studies THC seems to impair the immune response to cancer. As yet there is no evidence that THC or other cannabinoids have anticancer effects in humans. Third, Delta(9)-tetrahydrocannabinol (THC) may treat the symptoms and side-effects of cancer, and there is evidence that it and other cannabinoids may be useful adjuvant treatments that improve appetite, reduce nausea and vomiting, and alleviate moderate neuropathic pain in patients with cancer. The main challenge for the medical use of cannabinoids is the development of safe and effective methods of use that lead to therapeutic effects but that avoid adverse psychoactive effects. Furthermore, medical, legal, and regulatory obstacles hinder the smoking of cannabis for medical purposes. These very different uses of cannabinoids are in danger of being confused in public debate, especially in the USA where some advocates for the medical use of cannabinoids have argued for smoked cannabis rather than pharmaceutical cannabinoids. We review the available evidence on these three issues and consider their implications for policy.