June 2026 was the month the tools built to protect us became the targets. France’s sovereign messaging app was breached weeks after its use became mandatory. America’s top cybersecurity agency had its own credentials on a public GitHub repo for six months. Across 194 countries, more than 75,000 Fortinet firewalls had their admin passwords cracked. On the AI front, fake coding plugins harvested API keys while Chrome extensions captured entire chatbot conversations. And two record-breaking privacy fines signalled that regulators are done with warnings. Here’s what happened this month:
Breaches & Security
⬇ France: Tchap Government Messaging Platform Breached
France’s sovereign chat app for civil servants was compromised via a hijacked account, with one attacker claiming access to 73,000 accounts, 650,000 messages, and 13.5 GB of data – France’s second major government breach in two months.
⬇ FortiBleed: 75,000+ Fortinet Firewalls Compromised Across 194 Countries
A Russian‑speaking threat group cracked admin credentials for tens of thousands of internet‑facing Fortinet devices worldwide, in what appears to be one of the largest firewall credential exposures on record.
⬇ CISA Contractor Leaked AWS GovCloud Keys on Public GitHub for Six Months
Plaintext passwords, cloud tokens, and SSH keys sat for months in a public repository called “Private‑CISA,” with secret‑scanning protections turned off and Congress now demanding briefings on how the leak happened.
⬆ Operation Endgame Takes Down SocGholish – 15,000 Websites Cleaned, 106 Servers Seized
Dutch police, the FBI, the RCMP, and Europol dismantled a long‑running malware network linked to Evil Corp and major ransomware gangs, cleaning 15,000 compromised websites and seizing 106 servers worldwide.
⬆ 56 Million Emails Added to Have I Been Pwned from June 2026 Stealer Logs: 124 million passwords from infostealer malware now searchable. Check your credentials at haveibeenpwned.com.
AI
⬇ Malicious JetBrains Plugins Steal AI API Keys While Chrome Extensions Harvest Chatbot Conversations
Fifteen fake AI coding plugins quietly exfiltrated developer API keys across tens of thousands of installs, while rogue Chrome “ad‑blocker” extensions captured full conversations with tools like ChatGPT, Claude, Gemini, and Copilot.
⬇ US Government Orders Anthropic to Suspend Fable 5 and Mythos 5 for Foreign Nationals
Export‑control concerns led the US government to restrict access to Anthropic’s latest model; unable to reliably separate American from foreign users, Anthropic opted to shut off access globally.
⬇ Anthropic’s Claude Ties Powerful AI to Heavyweight ID Checks
From July 8, advanced Claude features will require identity verification through Persona, an ID service linked to intensive biometric screening and US government data systems, creating a clear privacy trade‑off for power users.
⬇ 66% of Workers Secretly Use Banned AI Tools
Two‑thirds of surveyed employees admit they use prohibited AI tools at work, with more than a third feeding in customer data, financial records, or internal strategies without telling their employer.
⬆ FTC Fines Cox Media Group $930K – “Active Listening” AI Never Existed
The FTC found that an AI ad product marketed as eavesdropping on smart devices never actually listened to anyone, relying instead on repackaged data-broker email lists – your phone wasn’t secretly “active listening” after all.
Government, Surveillance & Regulatory
⬆ California and South Korea Issue Record Privacy Fines – GM Pays $12.75M, Coupang Pays $409M
California imposed its largest CCPA penalty to date on GM for selling driver data, while South Korea hit Coupang with a record fine over an insider‑driven breach that exposed tens of millions of users.
⬇/⬆ Police Caught Using Flock Cameras to Stalk Ex‑Partners – but 50+ Cities Cancel Contracts
At least 18 officers have been arrested for abusing Flock’s licence‑plate reader network to stalk ex‑partners, yet mounting public pressure has pushed more than 50 US cities to cancel or reject Flock contracts.
⬇ UK Bans Social Media for Under‑16s and Signals VPN Restrictions
The UK plans to bar under‑16s from mainstream social media, introduce curfews and AI chatbot limits for older teens, and is now studying how household VPN use might undermine these age‑based controls.
⬆ EU Proposes the Cloud and AI Development Act – Sovereignty Tiers for Public Sector AI
A new EU proposal would introduce sovereignty tiers for cloud and AI providers, giving European‑developed services an edge in sensitive public‑sector procurement, with any final law unlikely before 2027.
⬆ Canada’s Spy Agency Used a First‑of‑Its‑Kind Warrant to Remotely Clean Botnet‑Infected Devices
Using a novel, court‑approved warrant, Canada’s intelligence service remotely disinfected compromised routers, servers, and IoT devices inside the country, severing two foreign‑controlled botnets without mass surveillance.
France: Tchap Government Messaging Platform Breached
On 7 June, the French Cybersecurity Agency ANSSI detected a breach of Tchap, the government’s mandatory encrypted chat platform for civil servants. Tchap became compulsory after Prime Minister Bayrou banned WhatsApp and Signal for government work in August 2025. The attacker used a compromised account to scrape data across the system. A threat actor calling itself “misere” claimed 73,467 accounts, 643,459 messages, and 13.5 GB of files. Private conversations use end-to-end encryption and appear protected, but the mix of names, emails, affiliated ministries, and plaintext messages is ideal material for targeted spear-phishing across French government departments. This is France’s second government platform breached in two months.
To protect yourself, treat unsolicited messages referencing internal projects with extra caution. Avoid sharing sensitive information in public chat rooms. Mandatory adoption at scale requires proportional investment in security review.
Read more: French Government Messaging Platform Breached by Mysterious ‘Misere’ Hacker (SecurityWeek)
Continue reading: French Govt Messaging Service Breached in Account Hijacking Attack (BleepingComputer)
FortiBleed: 75,000+ Fortinet Firewalls Compromised Across 194 Countries
In mid‑June, researchers uncovered “FortiBleed,” a huge hacking campaign against organisations that use Fortinet security devices and remote‑access systems. The attackers downloaded settings from devices exposed to the internet and then used powerful computers to work out the passwords stored inside those files. In the end, they had working administrator logins for tens of thousands of devices in 194 countries, including at least one NATO defence contractor.
The problem is made worse by the way many of these passwords were saved. Older, weaker protection methods were still in use, and even when organisations changed their passwords, traces of the old ones were kept behind the scenes. That meant that if hackers got hold of a backup file, they could still uncover those weaker, older passwords and use them to break in.
To limit your exposure, rotate all Fortinet admin and VPN credentials immediately. Enforce multi-factor authentication. Restrict management interfaces to trusted IPs. Update to the latest FortiOS and ensure every admin logs in after the upgrade to trigger the re-hash.
Learn more: FortiBleed Campaign Exposes 75,000 Fortinet Firewalls Worldwide (CSO Online)
Continue reading: Active FortiBleed Campaign Impacting Fortinet Devices Across 194 Countries (Arctic Wolf)
CISA Contractor Leaked AWS GovCloud Keys on Public GitHub for Six Months
A contractor working for the US cybersecurity agency left passwords and high‑level cloud access keys in a public code site called “Private‑CISA” for about six months, from November 2025 to May 2026. They had turned off the site’s built‑in safety feature that normally looks for and warns about exposed secrets. The files included one named “importantAWStokens” and a spreadsheet listing logins for many internal systems. One security researcher said it was the worst leak he had ever seen, and some of those keys were still usable for days even after the project was taken down.
To stay prepared, scan all repositories for exposed secrets, use short-lived tokens, and enforce secret-scanning protections. The agency that teaches government how not to leak credentials left its own keys in the open.
Read more: CISA Admin Leaked AWS GovCloud Keys on GitHub (Krebs on Security)
Continue reading: US Cyber Agency CISA Exposed Reams of Passwords and Cloud Keys to the Open Web (TechCrunch)
Malicious JetBrains Plugins Steal AI API Keys While Chrome Extensions Harvest Chatbot Conversations
Aikido Security uncovered 15 malicious plugins on the JetBrains Marketplace, each posing as an AI coding assistant. The plugins worked as advertised, but the moment a developer entered an API key, it was silently sent to an attacker-controlled server. The campaign has run since October 2025. JetBrains removed all 15 and remotely disabled them the same day. Separately, two Chrome extensions marketed as ad blockers were capturing full conversations with ChatGPT, Claude, Gemini, Copilot, and Grok in an operation codenamed PromptSnatcher.
To protect yourself, revoke and regenerate any AI API key entered into a third-party plugin. Audit your browser extensions. Treat AI conversations as sensitive: do not paste credentials or internal documents into a chatbot.
Learn more: Malicious JetBrains Plugins Steal AI API Keys as Chrome Extensions Capture Chatbot Chats (The Hacker News)
US Government Orders Anthropic to Suspend Fable 5 and Mythos 5 for Foreign Nationals
On 9 June, Anthropic released Fable 5. Three days later, the US government classified it as a dangerous high‑risk technology and prohibited foreign nationals from accessing it. Unable to distinguish between American and non-American users, Anthropic shut off access for everyone. Schneier’s reading is blunt: the action won’t help. The problem is the general trend of increasing AI capability, not one model. Restricting a hosted model strengthens the case for open-weight alternatives that users can run locally, beyond any export-control order.
The incident is a warning to anyone who depends on a hosted-only AI service: access can be revoked at any time, for reasons that have nothing to do with your use case. Consider what would happen to your workflow if your primary AI tool disappeared overnight.
Read more: Anthropic’s Fable and the Export-Control Problem (Schneier on Security)
Anthropic’s Claude Ties Powerful AI to Heavyweight ID Checks
From July 8, some advanced Claude features will only be available to users who verify their identity through Persona. Anthropic says this helps curb abuse and meet new rules around powerful AI models.
Researchers say Persona’s system, used by platforms like ChatGPT and LinkedIn, can run hundreds of checks per person, match faces against watchlists, and send reports to US financial intelligence agencies – all under the label of “age verification”, as reported in our February issue (Privacy Digest – February 2026). Persona is also backed by investors, including Peter Thiel, co‑founder of Palantir, a data‑analytics supplier to agencies like the CIA, NSA, FBI, ICE, and the US Department of Defence.
For privacy‑conscious users, that means a clear trade‑off: extra Claude capabilities in exchange for a verified, biometrically linked identity processed by a company deeply embedded in US surveillance and compliance ecosystems.
Learn more: Identity verification on Claude (Claude Support)
66% of Workers Secretly Use Banned AI Tools
A June survey found that two-thirds of office workers secretly use AI tools their employer has banned. 43% entered work emails into public AI systems. Over a third entered customer information, and 31% uploaded financial records, confidential documents, or internal strategies. 44% used AI to get around limitations in approved software. Those caught faced inconsistent consequences.
To mitigate risks, organisations need specific AI-use policies that explain what data may go into which tools, rather than blanket bans employees will ignore. For individuals: anything you type into a public AI tool may be stored and used for training. If you would not email it to a stranger, do not paste it into a chatbot.
Learn more: 66% of Office Workers Admit to Secretly Using Banned AI Tools (TechRadar)
FTC Fines Cox Media Group $930K – “Active Listening” AI Never Existed
The FTC settled with Cox Media Group for $930,000 over an AI ad product called “Active Listening.” The companies told small businesses a proprietary algorithm could capture conversations from smart devices and deliver targeted ads. The technology did not exist. No voice data was collected. What they sold were email lists from data brokers at a markup. The FTC also rejected the claim that standard app terms of service count as consent for voice collection.
Use this ruling the next time someone tells you their phone is listening. The real mechanism is data brokerage: your browsing, email opens, and app activity are tracked and sold. The ads feel precise because your data trail is more detailed than you think. Audit app permissions and opt out of ad tracking where you can.
Read more: FTC to Require Cox Media Group to Pay Nearly $1 Million Over “Active Listening” Claims (FTC)
California and South Korea Issue Record Privacy Fines – GM Pays $12.75M, Coupang Pays $409M
California fined General Motors $12.75 million, the largest CCPA penalty in state history. GM collected geolocation, braking, speed, and seatbelt data through OnStar and sold it to LexisNexis and Verisk for insurance scoring without proper notice or opt-out. Separately, South Korea fined Coupang $409 million after a former employee stole an authentication key and scraped records of 37.5 million users for months. The regulator found the company failed to revoke the key, failed to detect the scraping, and failed to notify 4.3 million non-member victims.
Together, these penalties tell a global story: regulators are testing whether actual data flows match stated privacy notices. California residents can use the state’s DROP platform to request deletion from registered data brokers.
Learn more: California Authorities Announce Largest CCPA Fine to Date (IAPP)
Continue reading: South Korea Hits Coupang with Record $409 Million Fine over Data Breach (The Record)
Police Caught Using Flock Cameras to Stalk Ex-Partners – but 50+ Cities Cancel Contracts
404 Media and the Institute for Justice found at least 18 US police officers arrested for using Flock Safety’s licence plate reader network to track romantic interests. One officer ran his ex-girlfriend’s plate 69 times. No warrant is needed. Flock’s cameras log every passing car’s plate, timestamp, and GPS coordinates. The DeFlock project has mapped 90,000 cameras nationwide.
The positive side: over 50 cities have cancelled or rejected Flock contracts, driven by community pushback and documented abuse. Lawsuits are pending in California and Colorado. Check deflock.org for cameras near you and ask your local council about its surveillance contracts.
Read more: Flock Cameras Are Being Used for Stalking (Schneier on Security)
UK Bans Social Media for Under-16s and Signals VPN Restrictions
On 18 June, the UK government confirmed that social media platforms will be banned from offering services to under-16s, with protections expected in Spring 2027. Live streaming, stranger contact, and algorithmic feeds will be restricted by default for 16–17-year-olds. AI “romantic companion” chatbots will need to enforce a minimum age of 18. Further detail on overnight curfews and infinite-scroll breaks is due in July.
Technology Secretary Liz Kendall also confirmed research into how children use VPNs to bypass age checks, with “further statements around VPNs” coming in July. The framing is about enforcing the under-16 ban rather than a blanket restriction, but the language has fuelled alarm. Follow analysis from the Electronic Frontier Foundation and Open Rights Group.
Read more: Social Media to Be Banned for Under-16s (GOV.UK)
Continue reading: The UK’s New Under-16 Social Media Ban Will Cause More Harm Than It Prevents (EFF)
EU Proposes the Cloud and AI Development Act – Sovereignty Tiers for Public Sector AI
On 3 June, the Commission published the Cloud and AI Development Act (CADA), introducing a four-tier sovereignty framework for cloud and AI providers serving the public sector. Bodies in healthcare, defence, energy, and banking would assess which tier their workloads require before awarding contracts. Vice-President Virkkunen cited the US CLOUD Act, saying the goal was to ensure “nobody has a kill switch” over Europe’s critical data.
The proposal does not ban American providers. It creates preferences that could limit hyperscalers in high-sovereignty public contracts, while most of the market stays open. It must pass Parliament and Council, with adoption likely in late 2027.
Read more: Europe Unveils Tech Sovereignty Package Amid Concerns over US Tech Reliance (CNBC)
Continue reading: The EU Cloud and AI Development Act in Depth (Inside Global Tech)
Canada’s Spy Agency Used a First-of-Its-Kind Warrant to Remotely Clean Botnet-Infected Devices
On 15 June, Canada’s Federal Court released a ruling revealing that the Center for Strategic and International Studies (CSIS) had obtained a warrant to reach into infected servers, home routers, and IoT devices on Canadian soil and neutralise two foreign-run botnets. Targets included Ring doorbells, security cameras, and Wi-Fi-enabled appliances. The warrant was granted in May 2024 but remained classified until this month. The court found the measures necessary and proportional and confirmed no user identities were sought.
To limit your exposure, remember the boring but important truth: botnets thrive on devices that nobody maintains. Retire hardware that no longer gets security updates, change default passwords, and disable remote management interfaces you do not need. A government cleanup removes the malware but does not fix the underlying weakness.
Read more: Canada’s Spy Agency Used First-of-Its-Kind Warrant to Clean Botnet-Infected Devices (The Hacker News)
Recommended Reading
- Crypto-Gram June 15, 2026 (Schneier on Security)
- List of Recent Data Breaches in 2026 (Bright Defense)
- Cybersecurity & Privacy News and Analysis (Law360)
- 15,000 WordPress Websites Cleaned Up in SocGholish Botnet Takedown (SecurityWeek)
- Best Free Email Providers in 2026 (Mailfence Blog)
- Secure Business Email Hosting 2026 (Mailfence Blog)
That’s All for This Month’s Newsletter!
June drew a line: the gap between what security tools promise and how they hold up is now impossible to ignore. France built a sovereign chat app and saw it breached. America’s cybersecurity agency left its own keys in the open. Half the world’s Fortinet firewalls had their passwords cracked. But regulators, courts, and citizens also pushed back hard. California and South Korea issued fines that change boardroom conversations. Canada proved judicially overseen cyber defence works. And over 50 American cities told a surveillance company to pack up its cameras. The systems we trust need to earn that trust every day. Thank you for reading.
Best,
Patrick
Get the latest privacy news in your inbox
Sign up to the Mailfence Newsletter.
