California's crackdown on election deepfakes arrives

A first piece of legislation, which will take effect immediately, i.e. before the November presidential election, effectively prohibits persons or groups from knowingly sharing certain deceptive elections. It is enforceable for 120 days before an election, but goes beyond that by remaining in force for 60 days after the election. A sign of concern about the spread of misinformation as votes are counted. Legitimate concern, we are told, thinking of what has happened in the past, not only recently, recalling the Bush vs. Gore saga, closed by the famous US Supreme Court ruling.

The other very significant legislation in the 'package' mentioned earlier has a name that is quite self-explanatory. "Law for the defence of democracy from the deception of deepfakes". It will take effect in January and will require social media platforms and other sites with more than one million users in California to tag or remove AI-generated deepfakes within 72 hours of receiving a complaint. If the platform fails to act, a judge can force them to do so.

In light of the content briefly described, the change from April Fools' Day to regulatory reality is hardly surprising.

The US elections are approaching, and in public opinion, but also at the political level, often with bipartisan orientations, a conviction is gaining ground. If it is true, as the Supreme Court has often repeated, thatfake news is in part protected by the First Amendment, precisely as a structural pre-condition for effective protection of the freedom of expression safeguarded by this constitutional provision, the American voter's right not to be defrauded at the founding moment of liberal democracies, that coinciding with the exercise of the right to vote, must be recognised.

There is, however, more, because, as mentioned, and this is a relevant passage, special attention is paid not only to the phase preceding the vote, but also to the phase immediately following it, the count, and also to the season that opens immediately after the count, that of the attribution of victory to one of the two sides and the regularity of the same. In other words, the basic idea, which necessarily also requires the active intervention of the large digital platforms, is that the American voter is not only defrauded at the moment when he exercises his right to vote, but also afterwards, with regard to the identification of the real winner, in a period that could extend, as we know, until the next election.

A final consideration. As I wrote some time ago on these pages, it is wrong to counterpose the Brussels effect, in the digital sphere, with a Washington effect, which does not exist, because the US Congress is blocked, as much with regard to the federal legislation, which has been pending for decades, on data protection, as with regard to the more recently proposed legislation on artificial intelligence, by a series of crossed vetoes.

Instead, one should speak of the California effect because that is now the American state from which the most 'European' regulatory initiatives originate, with a predictable domino effect, which however also means regulatory fragmentation among the 50 American states.

Meanwhile, more than something is changing, and His Majesty the First Amendment seems to have found its kryptonite in the fight against disinformation.