According to a disturbing BBC news story, South African drug addicts are stealing medication from HIV+ people and using it to get high:

'Whoonga' threat to South African HIV patients

"Whoonga" is, allegedly, the street name for efavirenz (aka Stocrin), one of the most popular antiretroviral drugs. The pills are apparantly crushed, mixed with marijuana, and smoked for its hallucinogenic effects.

This is not, in fact, a new story; Scientific American covered it 18 months ago and the BBC themselves did in 2008 (although they didn't name efavirenz.)

Edit 16.00 pm: In fact the picture is even messier than I first thought. Some sources, e.g. Wikipedia and the articles it links to, mostly from South Africa, suggest that "whoonga" is actually a 'brand' of heroin and that the antiretrovirals may not be the main ingredient, if they're an ingredient at all. If this is true, then the BBC article is misleading. Edit and see the Comments for more on this...

Why would an antiviral drug get you high? This is where things get rather mysterious. Efavirenz is known to enter the brain, unlike most other HIV drugs, and psychiatric side-effects including anxiety, depression, altered dreams, and even hallucinations are common in efavirenz use, especially with high doses (1,2,3), but they're usually mild and temporary. But what's the mechanism?

No-one knows, basically. Blank et al found that efavirenz causes a positive result on urine screening for benzodiazepines (like Valium). This makes sense given the chemical structure:
Efavirenz is not a benzodiazepine, because it doesn't have the defining diazepine ring (the one with two Ns). However, as you can see, it has a lot in common with certain benzos such as oxazepam and lorazepam.

However, while this might well explain why it confuses urine tests, it doesn't by itself go far to explaining the reported psychoactive effects. Oxazepam and lorazepam don't cause hallucinations or psychosis, and they reduce anxiety, rather than causing it.

They also found that efavirenz caused a false positive for THC, the active ingredient in marijuana; this was probably caused by the gluconuride metabolite. Could this metabolite have marijuana-like effects? No-one knows at present.

Beyond that there's been little research on the effects of efavirenz in the brain. This 2010 paper reviewed the literature and found almost nothing. There were some suggestions that it might affect inflammatory cytokines or creatine kinase, but these are not obvious candidates for the reported effects.

Could the liver be responsible, rather than the brain? Interestingly, the 2010 paper says that efavirenz inhibits three liver enzymes: CYPs 2C9, 2C19, and 3A4. All three are involved in the breakdown of THC, so, in theory, efavirenz might boost the effects of marijauna by this mechanism - but that wouldn't explain the psychiatric side effects seen in people who are taking the drug for HIV and don't smoke weed.

Drugs that cause hallucinations generally either agonize 5HT2A receptors or block NMDA receptors. Off the top of my head, I can't see any similarities between efavirenz and drugs that target those systems like LCD (5HT2A) or ketamine or PCP (NMDA), but I'm no chemist and anyway, structural similarity is not always a good guide to what drugs do.

If I were interested in working out what's going on with efavirenz, I'd start by looking at GABA, the neurotransmitter that's the target of benzos. Maybe the almost-a-benzodiazepine-but-not-quite structure means that it causes some unusual effects on GABA receptors? No-one knows at present. Then I'd move on to 5HT2A and NMDA receptors.

Finally, it's always possible that the users are just getting stoned on cannabis and mistakenly thinking that the efavirenz is making it better through the placebo effect. Stranger things have happened. If so, it would make the whole situation even more tragic than it already is.

Cavalcante GI, Capistrano VL, Cavalcante FS, Vasconcelos SM, Macêdo DS, Sousa FC, Woods DJ, & Fonteles MM (2010). Implications of efavirenz for neuropsychiatry: a review. The International journal of neuroscience, 120 (12), 739-45 PMID: 20964556