France's government deposited a ten-article bill in the Council of Ministers on July 9, 2026 — the projet de loi de cohésion républicaine par la lutte contre le racisme et l'antisémitisme — that would let the state order platforms to pull down racist and antisemitic content on the same fast administrative timeline currently reserved for child sexual abuse material and terrorist propaganda.
What the bill actually does
Article 7 inserts a new Article 6-2-2 into the 2004 digital-trust law (LCEN), extending Pharos — the interior ministry's online-hate reporting portal — into a takedown authority for racist and antisemitic speech. Hosting providers that receive a removal request from a designated police officer must act within 24 hours. Non-compliance carries up to six months' imprisonment and a €150,000 fine for individuals, rising to 2% of global annual turnover for corporate hosts. The bill also lets anti-racism associations join criminal proceedings as civil parties regardless of a case's severity, makes racist or antisemitic motive an aggravating circumstance across the board, and adds mandatory ineligibility for officials convicted of discriminatory offenses (Sénat bill text, deposited July 9, 2026; Élysée Council of Ministers readout). The government wants Senate examination from October and final adoption before year-end.
This is a second attempt. A narrower predecessor from MP Caroline Yadan, aimed only at antisemitism, collapsed in the National Assembly in April 2026 after opposition lawmakers argued it risked criminalizing criticism of Israeli government policy. Bergé's bill is broader — racism generally, not antisemitism alone — and routes the online-speech piece through the existing Pharos takedown architecture rather than new press-law offenses.
The case for it
The government isn't inventing a problem. Interior ministry data tallied by the anti-discrimination delegation DILCRAH show more than 16,000 racist, xenophobic, or antireligious offenses recorded in France in 2024, including 9,400 crimes and misdemeanors — an 11% rise on 2023 (DILCRAH, 2024 figures). Pharos itself logged roughly 20,000 discrimination-related reports in just the first half of 2025, out of 109,302 total reports across all categories (Pharos 2025 activity report). France has run an equivalent administrative takedown power for CSAM and terrorist apology since 2014 without collapsing into a censorship regime; extending the same mechanism, with the same 24-hour clock, to a second category of manifestly illegal speech is not a radical leap, and victims of targeted racist harassment have a real interest in content disappearing in hours rather than the months a criminal case can take.
Why the mechanism should worry a free-speech-minded reader
The honest problem is that "racist or antisemitic" is a much fuzzier line to draw at speed than "depicts child abuse" or "promotes a listed terrorist group." The Conseil d'État's own June 25, 2026 opinion on the bill flagged exactly this: it warned that provisions criminalizing non-public discriminatory speech could sweep in private conversations among family or friends alongside genuinely public posts, and it questioned whether converting several misdemeanor-level offenses into felonies survives proportionality review under French and European human-rights standards. On the takedown power specifically, the Council did not block it but insisted on safeguards — assessment by qualified personnel, not any officer, and prosecutor notification before a removal order issues (Conseil d'État opinion). Those caveats matter because the underlying design flaw of any 24-hour notice-and-takedown regime is well documented: platforms facing six-figure fines and criminal liability for guessing wrong will over-remove borderline content rather than risk it, especially when — as EFF has argued of automated moderation generally — enforcement speed increasingly means human judgment gets skipped entirely (EFF, "Automated Moderation Is Here to Stay," July 10, 2026). Amnesty International has separately criticized the bill for leaning on repression over education and prevention (Le Courrier de l'Atlas).
The proportionate path
None of this argues against giving Pharos more tools — the volume of reports alone justifies faster process than a years-long prosecution. But the Conseil d'État's safeguards are not optional extras; they're the difference between a targeted extension of an existing, judicially-tolerated mechanism and a vague speech offense enforced at platform scale under fine-driven incentives to delete first. As the bill moves to the Senate in October under the accelerated procedure the government has chosen — which compresses parliamentary debate — lawmakers should write the qualified-assessor and prosecutor-notification requirements directly into the statutory text rather than leaving them as guidance a future government could quietly drop. A 24-hour clock is defensible when the removal standard is genuinely unambiguous; it becomes a blunt instrument the moment platforms, not courts, are left to decide what counts as manifestly racist under threat of a €150,000 fine.