Designing an electronic voting system that is easy to use, efficient and secure may sound like an easy thing to do. And the pay-off — a democracy in which more people can participate and trust — seems desirable. But an academic analysis of three widely used systems in California has found monumental weaknesses in each of them. As a result, the state is slowing down its adoption of such systems until significant improvements are made. Others should exercise similar caution.

The study, commissioned by California's secretary of state, Debra Bowen, was led by computer scientists at the Berkeley and Davis campuses of the University of California. It found that the systems sold by three companies — Sequoia Voting Systems, Hart InterCivic and Diebold — had not been designed with security requirements in mind. And one particular deficit alarmed representatives of all political parties: the possibility that computer viruses could distort vote counts.

On 3 August, Bowen decertified the systems, which were already in use in counties where about half of the state's voters live. That means that in the primary elections next February, voters will return to paper ballots. Bowen has pledged to fully recertify the machines when they comply with a list of basic requirements: but the study authors question whether the software and hardware are amenable to ready repair. “They have serious security problems that will take years to fix,” says David Wagner, a study leader at the University of California, Berkeley.

This isn't the first time that specialists have warned against electronic voting systems. The Voting Technology Project, for example, a joint effort between the Massachusetts and California Institutes of Technology, highlighted their failings back in 2001 (see Nature 412, 258; 2001).

Yet the march of voting automation continues worldwide, often driven not by the public good but by election officials' desire for low staff costs and quick counts — as well as by the marketing machines of the systems' suppliers. Even in the United States, the Californian analysis is unlikely to make much of a difference in the many other states where the same electronic systems are being introduced. Verifiedvoting.org, a non-partisan lobby group that campaigns for reliable voting, says that although some secretaries of state are paying attention to the study, others — especially in the south and the midwest — don't seem to be interested.

There remains a body of public officials who seem to favour expediency and convenience over the democratic imperative of an accurate count. The firms that sell the systems have, meanwhile, argued that in the real world of elections, the systems will be overseen by election officials and candidates who would protect against the kind of disruptions identified in laboratory studies.

After the scandal that unfolded in Florida in the 2000 presidential election, when President George W. Bush eked out a narrow victory after prolonged legal arguments over disputed ballots in several counties, Congress passed a law that, among other things, helps to fund the replacement of existing, outmoded voting equipment. Now it is set to revisit the issue, with Senator Dianne Feinstein (Democrat, California) pledging to hold hearings that will pick up where the review in her own state left off. This may spur broader federal action to strengthen voting systems.

The consistent message from studies of electronic voting systems has been that the technology is often being implemented before it is ready to achieve the levels of security and reliability that voters are entitled to expect. Other jurisdictions worldwide should follow California's lead, consult with computer scientists, and act where necessary to stop this from happening.