‘The border is open’: a rumour sends 72,000 people to Ceuta, Google pulls a tool used to fake satellite images, and the EU's AI labels arrive

A rumour of an open border drives 72,000 people to Ceuta

On July 30, tens of thousands of people crossed from Morocco into Ceuta, a Spanish enclave on the North African coast, by land and by sea. By the time EU interior ministers met on Tuesday, the count stood at about 72,000, and about 75 people had died – most drowned, others were crushed in the crowd. Spain's interior minister put returns at around 70,000.

Spanish daily El País traced the surge to anonymous TikTok accounts claiming the border was open and no papers were needed. One widely shared video told viewers: “If you see these stairs, your dream will have come true.” Accounts posted where to buy wetsuits, flippers and diving goggles, and shared Google Maps screenshots pointing to Tarajal Beach on the Ceuta border – then the videos and the accounts behind them disappeared. The claims fed on a misread Supreme Court ruling about migrants intercepted at sea and on a regularisation programme that applies only to migrants already resident in Spain since January 1.

On Sunday, Morocco's interior ministry said the crossings into Ceuta and Melilla had been fuelled by misinformation on social media, human trafficking networks and misinterpretations of a Spanish court ruling. A 70-page reconstruction by security analyst Jose Maria Gil Garre traced the campaign from the July 8 court ruling to the mass crossing, identifying two Moroccan TikTok accounts with more than 400,000 combined followers and estimating over 11 million potential impressions. After Tuesday's emergency meeting, EU migration commissioner Magnus Brunner pointed the same way, blaming criminal smuggling networks and disinformation – “that's what we know for sure.”

Ukraine's foreign ministry said its analysts detected a pro-Russian network amplifying the images to destabilise Europe: more than 1,500 publications on outlets led by Pravda and RT, with Pravda's Spanish edition alone posting close to 300. Conspiracy theories framing the surge as a US–Israeli operation circulated widely; Euronews found little evidence for any of them. By Friday evening none of those who crossed had moved on to mainland Europe, the Commission said, and Italy still suspended passport-free travel with Spain for a month.

Google Earth got a button that invents satellite images – it lasted a day

On Thursday, July 30, Google added a “create image” feature to Google Earth's web version, powered by its Nano Banana 2 model. Users could generate photorealistic changes to real satellite views. Within hours, researchers were demonstrating the problem: Dutch OSINT researcher Henk van Ess produced refugees at the Mexican border, a nuclear plant in Iran, a crash in Amsterdam and a hospital with a bomb crater in Gaza. AFP journalists generated an explosion in Paris, an ISIS camp in Syria and a migrant caravan at the US–Mexico border.

The verification community reacted fast. GeoConfirmed, an open-source verification project, reported people were already using the tool to fabricate military positions and relabel buildings as schools, and assessed that its main practical use appeared to be creating and spreading disinformation. Ross Burley of the Centre for Information Resilience called the feature irresponsible.

Google pulled the feature on Friday, saying it would work on “stronger guardrails” and acknowledging that people place particular trust in Google Earth. The images carried watermarks marking them as AI-made and did not appear in the main Google Earth experience. The New York Times noted what is at stake with an example from 2014: after flight MH17 was shot down over Ukraine, Russia's defence ministry pushed altered satellite images to shift the blame, and investigators used Google Earth's historical imagery as the trusted reference that exposed the manipulation. That reference value is what researchers feared the new tool would undermine.

AI content in the EU now has to say so

Article 50 of the EU's AI Act took effect on Sunday, August 2. Providers of generative systems must mark AI-generated or manipulated content in a machine-readable way; anyone deploying a deepfake, or publishing AI-generated text on matters of public interest, must disclose it; and chatbots must tell users they are talking to a machine. The Commission frames the risks the rules target as “misinformation and manipulation at scale, fraud, impersonation.”

Non-compliance can draw fines of up to €15 million or 3 percent of worldwide annual turnover. The Commission adopted its final guidelines on July 20, and by the end of July about 190 organisations had signed the voluntary code of practice on transparency of AI-generated content. Systems already on the market get breathing room: the marking obligation for pre-existing generative models applies from December 2, and the AI Omnibus package agreed by EU co-legislators contemplates further transitional relief, so enforcement may start softer than the letter of the law suggests.

The transparency rules land in the same week the Act's rules on AI models became enforceable, a step Euronews describes as cementing Brussels' role as the world's most prominent AI regulator. The European AI Office now polices the obligations on general-purpose and frontier models – transparency about how a model was built, disclosure of copyrighted training content, risk mitigation – with most leading Western labs signed up to the compliance code of practice, Meta being the notable exception. Any decisive action is likely to draw friction with Washington, which has already attacked the EU's digital rulebook when it touches American companies.

Sweden writes its first playbook for psychological defence

On July 28 Sweden launched its first national strategy for psychological defence, built on four areas: countering improper information influence, strengthening the will to defend, preparing the media with secured channels to reach the public, and international cooperation within the EU and NATO. Civil Defence Minister Carl-Oskar Bohlin put the case in one line: “Our will to defend ourselves sits between our ears.”

The government issued new instructions to the Psychological Defence Agency to deepen its cooperation with the Armed Forces, the Military Intelligence Service and the Security Service. The timing is practical: Sweden holds parliamentary, regional and municipal elections on September 13.

The strategy caps a decade of build-up: Russian pressure on Sweden's information space dates back at least to 2015–16, when the state outlet Sputnik ran a Swedish-language service that published roughly 4,000 articles, many of them negative stories about the EU and NATO, per the Swedish Institute of International Affairs. Sweden's defence ministry answered by creating the Psychological Defence Agency in 2022, and in October 2024 every household – more than five million – received a preparedness booklet, “In Case of Crisis or War”, which tells Swedes that if the country is attacked, any claim it would surrender is false. The fabrications the agency deals with can be strange: this year Russian media spread a false story that Sweden's king, Carl XVI Gustaf, visited Ukraine but refused to shake President Zelensky's hand – supposedly out of lingering resentment that Peter the Great's Russia crushed the Swedish army at Poltava in 1709. The agency's director-general, Magnus Hjort, dismissed it in one line: “Absolutely nobody cares about this in Sweden.”