Foreign interference, TikTok, X, and elections: what the DSA actually changes (and what it doesn’t prove)

With France’s presidential election roughly eight months away, foreign interference is back in the feed: a fake “Blast” report, Storm-1516, TikTok, X, the DSA, talk of bans. The thread blends three different things — attributed disinformation campaigns, duties of very large platforms, and political proposals — until “the platform” and “the foreign state” sound like one threat.

This piece separates the layers. It is not legal advice and not a partisan editorial. Attributions (GRU, Storm-1516, and so on) are what services / prosecutors / the press say as of now: treat them as documented operational hypotheses, not a final court judgment.

“Interference” here ≠ classic espionage

In the 2026 French debate, ingérence / foreign interference usually means an attempt to shape a public debate or an election from outside — through false content, media impersonation, coordinated amplification — more than a spy photographing blueprints.

Different proof, different toolkit:

Idea	Classic espionage (schema)	Information interference (schema)
Goal	steal / sabotage secrets	smear, sow doubt, polarize
Trace	access, exfiltration, malware…	clone sites, deepfakes, accounts, ads
Useful proof	technical forensics + intel	attribution and evidence of impact
Visible actor	often invisible	platforms, algorithms, virality

An operation can be sophisticated and still barely seen. A local rumor can make more noise with no foreign link. That is the feed trap: treating “toxic content on X/TikTok” as the same fact as “a state flipped the vote.”

August 2026: the Glucksmann case (and the halo around it)

In early August 2026, MEP Raphaël Glucksmann (Place publique) — a sharp critic of Vladimir Putin and a possible presidential candidate — said France’s SGDSN (national defense and security secretariat) told him he was the target of an attack by Storm-1516, described in press and security sourcing as a pro-Russian information operation linked, per those same sources, to Russian military intelligence (GRU). To be clear: that attribution is the authorities’ and security sources’ reading as relayed by media — not a courtroom verdict.

The reported playbook (France 24 / AFP, franceinfo):

• a fake outlet copying the branding of French media Blast;
• forged documents and synthetic / faked voices (including deepfake);
• a story claiming his partner, journalist Léa Salamé, tried to “bribe” media for favorable coverage — denied by the journalists named and called a total lie by Glucksmann.

On August 5, on RTL (relayed by outlets including BFMTV), Glucksmann urged authorities not to have a “trembling hand” toward Elon Musk / X and the Chinese Communist Party / TikTok, and to force big platforms to follow European rules — or face sanctions. Useful angle for this decode: he is pointing at the regulatory lever, not only Moscow.

On August 6, the Paris prosecutor’s office opened an inquiry (Ouest-France / AFP) into whether, among other things, faked images were produced to serve a foreign power’s interests. Opening an inquiry ≠ proven guilt. The case will move; the media facts are already in the debate.

Attal, Philippe: same family of attacks, according to the press

Several outlets (Ouest-France, Le Figaro, and others) report that Édouard Philippe and Gabriel Attal were also targeted in the same summer wave by destabilization ops attributed to pro-Russian networks (invented health claims, and similar). Here too, we relay the press / prosecutor framing, not a complete judicial map. The decode point: this is not “one candidate” — it is a pattern testing multiple targets.

Tondelier and the idea of banning X during elections

In an interview with Libération (early August 2026), Marine Tondelier argued for threatening platforms such as X with a ban during elections if interference is established beforehand — and for applying the DSA framework. That is a political proposal, not a measure in force, and not a done deal.

Decode without caricature:

• the debate is about deterrence (a credible threat) vs a hard cut (an actual shutdown);
• under EU / French law, a broad suspension of a service remains an exceptional tool, tightly framed (proportionality, procedure, often judicial control) — media fact-checks (TF1 Info and others) stress how hard that is in practice;
• conflating “sanction a VLOP under the DSA” with “cut off X for a month” mixes fines / orders / duties with national access blocking.

Saying “X will be banned” would be false. Saying “a candidate proposes a conditional ban threat” is accurate.

Romania 2024 / TikTok / Georgescu: the cautionary case

The Romanian 2024 case keeps coming back: Călin Georgescu, a far-right candidate previously marginal in polling, led the first round in November 2024 amid Romanian authorities’ claims of illicit promotion, notably on TikTok, with suspicions of Russian influence. Romania’s Constitutional Court then annulled the vote, which was rerun in spring 2025. Handle with a strong caveat: each country has its own election law, evidence, and constitutional court, and the causal link between the online campaign and the result is still contested. This is not a kit for “what will happen in France.”

What the precedent actually helps say:

1. platforms can become the stage for an electoral legitimacy crisis;
2. annulling a vote is a heavy, rare, contested institutional act;
3. “not enough visible sanctions against TikTok” (an argument reused in the French debate) ≠ “the platform alone caused the result.”

Useful as a mechanism warning; misleading as a prophecy.

DSA: what platforms must do

The Digital Services Act has applied in full since February 2024 (Vie publique). It puts duties on online intermediaries, with heavier obligations for Very Large Online Platforms (VLOPs) and search engines (VLOSEs) — typically ≥ 45 million monthly EU users. TikTok and X are among designated services.

For VLOPs, the “systemic risk” core (including large-scale manipulation / disinformation) includes, among other things:

• assessing and mitigating systemic risks;
• audits, transparency, cooperation with authorities;
• for the most serious breaches, fines of up to 6% of worldwide annual turnover — a statutory ceiling, not an automatic fine for every fake post.

From mid-2025 (effect July 1, 2025), the Code of Practice on Disinformation (strengthened in 2022) was integrated as a code of conduct under the DSA (Commission + Board opinions, Feb. 2025). For VLOP/VLOSE signatories, living up to those commitments becomes a benchmark for DSA compliance on disinformation risk — still co-regulation, not line-by-line state censorship.

In France, Arcom is the Digital Services Coordinator; the European Commission keeps exclusive competence over certain enhanced VLOP duties.

What states can prove — vs what platforms must prevent

Two different clocks:

Question	Platforms (DSA / Code)	State / courts / intelligence
Main duty	prevent / mitigate systemic risks, remove illegal content, cooperate	investigate, attribute, prosecute
Useful proof	metrics, process, audits, response times	attribution, intent, impact
Success looks like	less spread, more transparency	established facts, possible criminal sanctions
Common failure	“we have a policy” with no measurable effect	attribution without proof of electoral effect

A prosecutor can open a case on faked images serving foreign interests before the Commission has ruled on a DSA file about the same episode. Conversely, a platform can face DSA pressure over a systemic risk even if no public “GRU” attribution exists yet.

The feed wants one sentence (“Musk / TikTok / Putin”). The useful mechanism is dryer: content × distribution × attribution × law.

Mini glossary

• Foreign interference (2026 debate): outside influence on debate / elections, often via information ops.
• Storm-1516: pro-Russian disinformation modus operandi cited by security services and press; handle the attribution with care.
• VLOP: very large online platform under the DSA (≥ ~45M monthly EU users).
• Systemic risk: large-scale risk (including public-debate manipulation) that VLOPs must assess and mitigate.
• Code of Conduct on Disinformation: commitments (ex–Code of Practice) integrated into the DSA framework mid-2025 as a benchmark.

Bottom line

Summer 2026 put three threads on one table: attributed disinformation ops (Glucksmann / fake Blast / prosecutor inquiry; multiple targets per the press), platforms (X, TikTok) accused of being both channel and lever, and a DSA framework that treats disinformation as a systemic risk for VLOPs — with fines that can reach the regulation’s statutory maximum of 6% of global turnover, and a Code of Conduct integrated since mid-2025. Tondelier’s idea of banning X during elections remains a conditional political threat, not governing law. The Romanian precedent shows a platforms × ballot crisis can turn institutional — without supplying a copy-paste script for 2027. Decoding means separating what is attributed, what is proven, and what regulation can compel.

Going further

• France 24 / AFP — operation targeting Glucksmann: timeline and SGDSN / Storm-1516 attribution.
• franceinfo — “Everything in it is false”: fake Blast detail and denials.
• Ouest-France — Paris prosecutor inquiry (Aug. 6): case opening and criminal framing mentioned.
• Vie publique — DSA: obligations and VLOP logic.
• Commission — Code of Practice → DSA: Feb. 2025 endorsement / July 2025 effect.
• franceinfo — Tondelier / Musk: conditional ban proposal and reaction.