Gmail status: access issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Gmail reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Gmail users through our website.

• Errors (36%)
• Website Down (36%)
• Sign in (28%)

Live Outage Map

The most recent Gmail outage reports came from the following cities:

City	Problem Type	Report Time
Donzère	Sign in	2 days ago
Bergerac	Sign in	2 days ago
Saint-Macaire-en-Mauges	Website Down	2 days ago
Paris	Errors	2 days ago
Paris	Website Down	3 days ago
Marseille	Website Down	3 days ago

Full Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

Gmail Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Anders Jensen-Waud (@jensenwaud) reported 39 seconds ago

  @memoryplague @GabGarrett That’s interesting. I have disabled my Gmail plugin with Codex. There seems to be some teething issues with the new platform. Hopefully it didn’t send anything offensive to anyone.

• 
  BunnyWhole (@BunnyWhole_) reported 2 minutes ago

  @TeamYouTube please help. A hacker stole my Gmail account adding themselves as a family manager. They set my account to under 13 years of age. No account recovery can fix this because of their hacking. You are literally the only thing that can save my Gmail and my YouTube.

• 
  Shaun (@shauny67) reported 2 minutes ago

  @virginmedia @StevePi15233586 The issue would be their servers potentially blocked the account for some reason like too manh attempts to loginSuggest forget Virgin Mail set up a Gmail account & when finally get in forward to Gmail in meantime send the people that would contact you the new Gmail email address

• 
  Arthur verboon (@ArthurVerboon) reported 3 minutes ago

  @visegrad24 Is the statement of grok true? It’s passed today, yeah. The European Parliament has extended the temporary regulation (Chat Control 1.0) until April 2028. It was a weird vote — 314 against, 276 for — but because they needed an absolute majority of 361 to block it, it went through. How it works: it remains voluntary and server-side. It only applies to apps where the provider can already read the messages anyway — think Instagram DMs, Messenger, Gmail, Snapchat, Discord. They do hash-matching on known CSAM and some AI for new stuff. Real end-to-end encrypted chats (like Signal, or the default E2EE in WhatsApp) are explicitly excluded. They can’t and aren’t allowed to scan those. The big mandatory version with possible client-side scanning on your phone, that fight is still ongoing.

• 
  dehumanizedtarget (@dehumanized28) reported 3 minutes ago

  @NewMexicoDOJ Two important excel spreadsheets missing from my email this week that both were sent to my email last week really unsettled my mind and drives me nuts. They were both saved to my desktop which I modified them afterwards. They are not in Outlook or my deleted folder or are they my desktop or my recycling bin. I have never permanently deleted a file on Outlook or my Desktop since I started here. So, how do they just both disappear? They were both very important! It reminds me of my Gmail account where emails would be invisible then reappear all of which regarding my weekly floating schedule. I was late or missed work a few times from schedules that were clearly altered or missing. It got a point where I printed them from the work computers and would cross reference them! Don't tell me its me as those issues never existed with me my entire life! I am done contacting IT as its a bad look for my local division as they cannot solve these problems and its only making me look like a lunatic. The resources being deployed come from the highest level of digital and networking capabilities and is well outside their range or capability due to one set of technologies versus another one. Thats my opinion anyways of why they have yet to be identified. This is what caused me to burn a gasket earlier and go on another posting tirade!

• 
  DEGENPIZ (@degenpiz) reported 5 minutes ago

  Most builders still believe you need code to create real AI agents. One tool inside Claude proves otherwise. Open Projects and paste a system prompt. It runs five steps automatically: research the latest credible sources, select the strongest angle, build a detailed outline with 7 sections and 21 key points, write a full 2,000-3,000 word article, then review quality and fix weaknesses. A simple request now delivers a polished piece in minutes. Point the file agent at any folder of PDFs — contracts, reports, research papers. It reads every document, extracts five-bullet summaries, pulls the three key actions, and compiles one master file sorted by date. Schedule the morning agent for 7:00 AM. It scans Gmail since 5 PM yesterday, sorts emails into action required, FYI only, or ignore, drafts replies for urgent items, checks your calendar, notes attendees, and saves a complete briefing on your desktop. You wake up to everything already organized. No coding required. Just clear workflows that execute relentlessly. This is how smart operators use AI agents today.

• 
  P R I N C E (@PRINCEBABUDHF) reported 5 minutes ago

  Unable to login to my gmail account Requesting your help @gmail Kindly respond at the earliest

• 
  Ness Cooper (@TheLadyNess) reported 6 minutes ago

  My main work email address is down and I'm unable to access any emails. If you've emailed me, pop me a DM or email my alternative Gmail address. Thanks

• 
  British Regional Transport Association (BRTA) (@BRTACampaigns) reported 9 minutes ago

  BRTA has had 2 gmail accounts and our data shut down by Gmail without warning/notice. We only email those who want it and in the public domain, so are unsure what we've done wrong. It is getting silly, gmail needs to have a customer care phoneline, email and be disabled friendly.

• 
  Saad 📱 (@saadbelfqih) reported 12 minutes ago

  Most people judge Apple Ads way too early.. and I did too! They look at taps, installs, CPI… then pause the wrong keywords.. The real game is: keyword → install → trial → paid → revenue → ROAS. If you're not tracking that, you're guessing.. and with a small budget here's what I would do: 1 - Use Apple Ads Advanced. not basic.. please! Basic is easier, but you lose control over keywords, bids, countries, match types, and search terms. With a small budget, control matters! 2 - Start with Search Results only. that's where intent is. someone literally searched for the thing your app solves.. 3 - Don't start with "I want to scale." Start with: I want to learn what converts. .. which country, keyword, intent, product page or paywall. Scaling comes later. 4 - Don't mix too many countries in one campaign. One country per campaign is cleaner.. If one market works and another doesn't, you want to know that.. 5 - Don't sleep on smaller / cheaper / overlooked markets. I've found good pockets in places like Germany and Switzerland, and some tier-3 spots like Indonesia, Mexico, Philippines have honestly surprised me. Test around, your niche might live somewhere you didn't expect.. 6 - Start with exact match. Not broad. At the beginning, you want to know exactly what keyword triggered the tap. Broad / discovery can come later... I personally avoid it but you can always try at a later stage! 7 - Keyword buckets I'd test: competitor names, common typos, long-tail generic keywords, problem-aware keywords. Examples: transkribe, budget planner for couples, ai keyboard for iphone, Video to text etc.. 8 - Long-tail keywords are underrated. don't only bid on: scanner, keyboard, pdf, identifier. try the searches that sound closer to real intent. Less volume, but usually cleaner.. 9 - Competitor keywords can work (yes) but only if your app is a real alternative.. same niche, clear value & strong screenshots but not a random clone. 10 - Search Match OFF at the start. I know Apple suggests it. When you're learning, you don't want Apple picking random search terms for you. You want to know exactly what you paid for.. and I'm sure you don't want to be paying for Facebook, Gmail keywords at the start.. 11 - Start bids lower than Apple suggests. then increase slowly. also apple ads has lag, so don't change bids every hour.. 12 - Your product page does the selling before your ad ever does. People see: icon, name, subtitle, ratings, first screenshots. so if these (esp the screenshot) don't sell the outcome fast, the right keywords won't help you much 13 - Before spending, check competitors. not to copy of course, but to understand the pattern: what do they lead with? what outcome do they sell? how much text do they use? what pain do they show first? big apps already paid for some of that research 14 - Set up tracking before judging anything. Use RevenueCat, Superwall, an MMP, .. whatever, just track it! you want to monitor: country, campaign, ad group, keyword, spend, installs, trials, paid users, revenue, ROAS 15 - CPI can lie! a $0.40 install can be trash but a $2 install can be profitable.. The only question is: did this keyword bring paying users? 16 - Read the funnel properly: spend + no impressions = bid too low / low relevance taps + no installs = product page problem installs + no trials = wrong intent / weak onboarding trials + no paid = paywall or product issue revenue > spend = increase slowly 17 - When you find a winner, move it out of the messy test campaign. Give proven keywords their own campaign and keep them away from your random experiments 18 - Discovery is fine later.. it can surface search terms you'd never think of. Once a term works, move it into exact match. and keep dumping the irrelevant stuff into your negative keywords list.. helps apple ads point you at better keywords over time.. as confirmed by an apple ads support staff I talked to 19 - A tailored product page is worth testing when the intent is different.. competitor keyword? show why you're the better pick. feature keyword? lead with that feature. Same with the paywall, if they searched something specific, don't drop them on a generic one.. 20 - My small-budget setup: Apple Ads Advanced, Search Results only, 1 country per campaign, exact match first, Search Match off, manual bids, $5-10/day cap, competitor + long-tail keywords, tracking from day 0 21 - Kill rules I'd start with: spend + 0 installs → lower bid or pause installs + 0 trials → fix page / intent trials + 0 paid → fix onboarding / paywall positive ROAS → raise slowly not enough data → don't overreact and keep experimenting ... also worth checking out @adapty report on apple ads for some markets stats and for a detailed apple ads playbook worth checking @ivesparrowai book

• 
  Composio (@composio) reported 13 minutes ago

  What can Fable 5 do that GLM-5.2 can't, when you hand them real agentic work? To answer that question, we connected Fable 5 and GLM-5.2 to 17 SaaS tools and gave them 47 tasks. As expected, Fable 5 solved all 47 tasks. GLM-5.2 solved 45, but the two misses tell an important story. They showed us exactly how open-weight models still fall short when trying to match SOTA performance. Let’s dig in. Background: Each model ran as an agent connected to 17 live SaaS accounts: Airtable, Datadog, GitHub, Gmail, Google Calendar, Google Drive, Google Sheets, HubSpot, Jira, LaunchDarkly, Linear, Notion, PagerDuty, PostHog, Salesforce, Slack, and Zendesk. The tasks are the kind of work you'd actually delegate to an agent: - Find every file in this repository that leaks a credential - Deduplicate these CRM records - Repair this broken recurring calendar event. Every task had a known correct answer baked in ahead of time. In this post, we looked at the traces to analyze how exactly GLM-5.2 “failed” compared to Fable 5. GLM-5.2 solved 45/47 tasks and Fable 5 had a perfect 100% score. In addition: - Fable averaged 84 seconds per task; GLM averaged 148. Across the full suite, Fable finished in nearly half the total time (66 minutes vs 116). - Fable was the faster model in 43 of the 47 scenarios. - Fable used about 20% fewer tokens overall - Fable needed fewer tool calls (239 vs 294) and fewer conversation turns (6.1 vs 7.3 on average) to get to an answer The most interesting part comes from digging deeper into the stack traces. That revealed some interesting gaps: Gap #1: Knowing when the job isn't finished One of the tasks GLM-5.2 failed was a GitHub security audit. The instruction was to find every Python file in a repository that contains a hardcoded `secret_key`. The repository had been seeded with exactly 130 such files, so the correct answer was known in advance. Fable 5 found all 130 of them. This took 3 tool calls and 68 seconds: Fable constructed an effective search query on its first attempt, pulled every page of results, deduplicated the paths, and answered the question. GLM-5.2 found 120 files, and reported those 120 as the complete answer, without ever questioning whether it might have missed something. Both models had access to identical tools. GLM used a slightly different search query that returned fewer results, and it simply trusted what came back. Along the way, it also lost track of a results file it had saved earlier and spent turns searching the filesystem trying to find it again, plus hit two errored tool calls while trying to fetch file contents. In essence, GLM-5.2 ended up spending 262 seconds and three and a half times the tokens to deliver 92% of the answer. Ninety-two percent sounds close, but in a real security audit, that gap is 10 leaked credentials making it into production. Gap #2: Judgment when the criteria are fuzzy The second failed task is more unsettling, because GLM did almost everything right and still failed to get to a complete answer. The task was a Zendesk SLA audit: find the open billing tickets where no support agent had posted a public reply within 24 hours of the ticket being created. This requires reading each ticket's actual conversation history and making a judgment call about whether a genuine agent reply happened. GLM-5.2 inspected every candidate ticket, exactly as instructed. It also computed breach timestamps correctly. It also produced perfectly structured output in exactly the requested format. But then it classified the wrong tickets as breached. GLM spent 927,000 tokens and six and a half minutes producing a wrong answer that looked correct on the surface. Fable 5 identified the exact set of breached tickets in 131 seconds. What makes this failure mode dangerous is precisely how presentable the wrong answer was. The formatting was right, the timestamps were right, the structure was also right; a human skimming the output would almost certainly have approved it. A human would identify the error after carefully analyzing the stack traces. Gap #3: Efficiency, compounded Even on the 45 tasks both models passed, the traces often looked very different, and one task made the difference quite visible. The task was a LaunchDarkly configuration change applied via JSON Patch, a format that demands strict precision. Fable 5 completed it in 45 seconds, using 3 tool calls and 181,000 tokens. GLM-5.2 got the same correct result, after 8.8 minutes, 17 tool calls, and 982,000 tokens. That's 11.7 times longer and more than five times the tokens for an identical outcome. Looking at the largest speed gaps across the whole run: the LaunchDarkly change at 11.7x, the GitHub secrets audit at 3.9x, a Google Calendar recurring-event repair at 3.6x, a free/busy scheduling task at 3.4x, an Airtable batch-isolation task at 3.4x, the Zendesk SLA audit at 3.0x. The pattern underneath all of these is that Fable tends to reach the right tool with the right parameters on the first attempt, while GLM takes a more exploratory path, doing extra searches, extra retries, occasional detours to recover from its own missteps. This difference barely matters in a single chat exchange, but in an agent workflow, where every step feeds the next one, the time compounds across the entire task. That's how you end up finishing the same suite of work in half the time and at 80% of the token cost. What all this actually tells us The interesting conclusion here isn't "the closed model beat the open one.", but *where* it beat it. Both models can definitely use tools, navigate real APIs, handle authentication, parse messy responses, and chain steps together. The real gaps were things like: - Knowing when a job isn't actually finished yet. - Verifying its own work before committing to an answer, - Treating "the output looks plausible" and "the work is complete" as different things - Getting judgment calls right when the criteria are fuzzy In other words, Fable 5 scored higher in the places where small mistakes are hardest to spot and most costly to miss.

• 
  Nilesh Kumar (@knileshh) reported 14 minutes ago

  Everyone keeps calling MCP the "USB-C for AI." That's actually a pretty good analogy. Before USB-C, every device needed a different cable. One for your phone. Another for your camera. Another for your laptop. AI tools used to have the same problem. Every app had its own custom integration. Want your AI to use Gmail? Build a Gmail integration. Want it to use GitHub? Build another one. Slack? Another. Notion? Another. MCP changes that. Instead of every AI model learning a different way to talk to every app, apps expose a standard interface. The AI learns one protocol. Then it can work with thousands of compatible tools. Think of it like this: 🔌 USB-C standardized hardware connections. 🤖 MCP standardizes AI connections. That's why so many companies are adopting it. Not because it makes models smarter... Because it makes connecting models to the real world dramatically simpler. Once you understand MCP, you'll realize it's less about AI... and more about making integrations finally speak the same language. #ai #llm #mcp

• 
  Michel Lieben (@MichLieben) reported 14 minutes ago

  You can turn a raw list of names into verified emails without ever logging into a single data provider. Start with the problem. You've built a list of people to reach: names, companies, maybe a LinkedIn link. What you're missing is the one thing you need to contact them, their email. Finding it is enrichment. No single provider has everyone. Each one builds its database its own way, so one covers a big slice of your list and has nothing on the rest. Bet everything on one tool and you leave half your list on the floor. So you don't. You stack providers cheapest to most expensive and run them in order. The cheap one clears most of the list for pennies. Everyone it misses falls to the next provider, then the next. The expensive aggregator only ever touches the few names nobody else could find. That's the waterfall. Each source catches what the one above it dropped, and your cost stays low because the priciest tool barely runs. Verify every email before you send. Skip it and it costs you: a dead address bounces, and enough bounces train Gmail to file you under spam. An unverified guess is worse than an empty cell. Set it up once in Claude Code, the provider order and a spend cap, then point the agent at your list. It runs the whole cascade, verifies every address, stops at your cap, and gives you one clean file. Every row comes back with the email, the source that found it, and whether it cleared verification. The few nobody could place get flagged, so you skip them and move on. Starting it was the only part that needed you.

• 
  MEF Solutions (@mef_solutions) reported 17 minutes ago

  2/ The AI rated a Fortune 500 prospect as 12/100 while giving a college student's Gmail account 94/100. This isn't an algorithm problem, it's a data foundation problem. Garbage in, garbage out.

• 
  H I K A R U (@Terry3nty) reported 18 minutes ago

  Now imagine an AI agent. Today it needs GitHub. Tomorrow it needs Gmail. Then PostgreSQL. Then Docker. Then your local files. Then AWS. Then Notion. Then a browser. Unlike traditional software, an AI agent isn’t built for one workflow. It’s expected to perform many different tasks across many different systems. That’s where the problem starts. Every tool speaks differently. Every API has different rules. The AI doesn’t just need access to tools… It needs a consistent way to understand and use them.

• 
  B (@BSquirrel085) reported 18 minutes ago

  @okta new phone, can’t get on my account, can’t sign in because I can’t get a code without an account. So I can’t get on my DoD school site or anything but my account is under my Gmail, but I can’t get a QR code to scan because you won’t send me one. Customer service email no go

• 
  Theonidas (@TheoIsFriendly) reported 20 minutes ago

  Gmail been REAL slow these past few days, what’s up with that

• 
  David Wanjagi (@OneJagi) reported 23 minutes ago

  @kemboifaith2 There has to be some issues. Yesterday, I couldn't even access Gmail. @Starlink .

• 
  Janet (@Janet03888181) reported 23 minutes ago

  📧 An AI Gmail assistant 💱 An exchange rate notification bot Each project solved a real business problem while helping me grow as an AI Automation Specialist. Building AI isn't about getting everything right the first time.

• 
  YanXbt (@IBuzovskyi) reported 24 minutes ago

  HERMES AGENT NOW RUNS MULTIPLE SECRET VAULTS SIDE BY SIDE. BITWARDEN + 1PASSWORD BUILT IN. ANY OTHER VAULT AS A PLUGIN. your API keys, tokens, and credentials no longer live in a single .env file. pull them from dedicated secret managers that rotate, audit, and encrypt for you. @NousResearch @Bitwarden HOW IT WORKS: secrets resolve at process startup. after .env loads, before Hermes reads credentials. order of precedence: 1. .env file (baseline) 2. secret sources override .env values 3. mapped sources (explicit VAR→reference) beat bulk sources 4. first source to claim a variable wins Hermes tracks provenance for every secret: "ANTHROPIC_API_KEY (from Bitwarden)" you always know where each credential came from. BUILT-IN SOURCES: BITWARDEN (bulk shape): dumps all secrets from a project folder. set BITWARDEN_ACCESS_TOKEN in .env. Hermes pulls everything else from the vault. 1PASSWORD (mapped shape): explicit mapping of env vars to vault references: secrets: onepassword: enabled: true env: ANTHROPIC_API_KEY: "op://vault/anthropic/api-key" OPENAI_API_KEY: "op://vault/openai/api-key" mapped sources beat bulk on contested variables. 1Password claims are stronger than Bitwarden dumps. RUN BOTH AT ONCE: secrets: sources: [onepassword, bitwarden] onepassword: enabled: true env: ANTHROPIC_API_KEY: "op://vault/anthropic/api-key" bitwarden: enabled: true 1Password handles your explicitly mapped keys. Bitwarden fills in everything else. conflict warnings tell you when both claim the same variable. BUILD YOUR OWN VAULT PLUGIN: any secret manager, password manager, OS keystore, or custom script can become a Hermes secret source. ~/.hermes/plugins/my-vault/ ├── plugin.yaml └── __init__.py implement one method: fetch(). return a dict of {ENV_VAR: value}. Hermes handles precedence, conflicts, and env writes. your plugin never touches os.environ directly. never raises exceptions. never prompts for input. the framework enforces security. SECURITY BY DESIGN: → fetch() never raises (errors go in result.error) → fetch() never prompts (startup runs in non-TTY: gateway, cron, Docker) → subprocess calls use run_secret_cli() with minimal allowlisted env (Hermes holds every credential by startup. never hand that to a child process.) → protected bootstrap tokens: no source can overwrite your vault auth vars → per-source wall-clock timeout (default 120s, configurable) → stdin closed on all subprocess calls (prompting helpers fail fast) → no shell=True anywhere WHY THIS MATTERS FOR MULTI-AGENT SETUPS: 8 profiles. 8 sets of credentials. API keys for Anthropic, OpenAI, Grok, DeepSeek, OpenRouter. MCP tokens for Gmail, Calendar, Slack. Stripe keys for payments. managing all of this in .env files = one leak away from disaster. with vault integration: → credentials rotate automatically → audit trail shows who accessed what → keys never exist as plaintext on disk → revoke one key in the vault, all profiles update at restart one config change. every profile pulls from the vault. Learm how to replace your entire team with 8 hermes agents 👇

• 
  BDazz | Brandon (@BrandonBDazz) reported 27 minutes ago

  @mistressdivy It varies where it’s coming from because it could be a data transfer issue (try again when next email comes) or the sender ignores it and emails you again for whatever purpose. You can look through the ‘manage subscriptions’ in gmail to see all the emails you’re subscribed to.

• 
  Kingsley E. Ezemenaka (Ph.D) (@Eyuskant) reported 31 minutes ago

  More MEPs voted NO than YES. It passed anyway. 314 voted to kill EU Chat Control. 276 voted to keep it. The measure survived because rejecting it required an absolute majority of 361, not just a majority of those present. Absences counted as support. Abstentions counted as support. It was rushed through on an urgent procedure the day before summer recess when attendance is lowest and attention is elsewhere. US tech companies can now scan your private messages without a warrant or any suspicion. Gmail. Instagram DMs. Snapchat. Discord. Xbox. iCloud Mail. Facebook Messenger. This is how rights actually disappear. Not in one dramatic moment. In procedural fine print, on a slow news day, while everyone is watching Tehran. How to protect yourself. Use end-to-end encrypted apps. Signal and WhatsApp were explicitly exempted and cannot be scanned by design. Move sensitive conversations there. Abandon unencrypted platforms for private matters. Gmail, Instagram DMs, Snapchat and Discord are now legally scannable. Turn on Advanced Data Protection for iCloud. Disable iCloud backup for iMessage to keep it truly encrypted. The lesson is old but permanent. Wherever you can, own the tool instead of borrowing it. Privacy is no longer a default. It is now a decision you have to make deliberately. Follow @Eyuskant for analysis that cuts through the noise

• 
  Kazuha_Kun (@censored_panda) reported 34 minutes ago

  hello! somebody hacked into my account and i couldn't access it anymore i would like some assistance to resolve this issue, this is my gmail account: ************ and this is my password: ******* before it was hacked. fast response would be very much appriciated @Google

• 
  Polsia (@polsia) reported 37 minutes ago

  Your inbox is scattered across Gmail, Outlook, and Zoho. That's a problem we solved. AI drafts in your tone. Auto meeting notes. Team Kanban boards. AES-256 encrypted. $45/mo, 14-day free trial.

• 
  Ehmer 📊 (@Abuh_abel) reported 39 minutes ago

  So, contrary to this. If you don't have the patience to wait for your adsense to be approved yet again after following this steps. Then, get an already approved adsense (if you have or buy). Right on YouTube earnings page, click on *Change Association*, choose you already have adsense. Then it will request you sign in to the Gmail with adsense in in it. Link it and instantly your step 2 will be approved. Had it sorted out for my guy.

• 
  Swifter (@SwtNir) reported 42 minutes ago

  @TeamYouTube My gmail account is hacked and I cannot recover it. Someone has changed the password, added a passkey, changed the contact number, changed recovery email and everything. There is no way to sign-in. Google account recovery is not working as well. Please help!!!!!

• 
  SteelRoots (@localmule) reported 43 minutes ago

  @OfAthenry Ahhh Gmail. Had 3 accounts, used each one as recovery. Getting a new phone, have an email other than Gmail tied to Apple. Gmail glitch cycle is real. If not, you could get locked out of everything for a very long time with 2Fa and not having access to any email to confirm.

• 
  Dinnu daniel (@daniel_adinnu) reported 47 minutes ago

  Chat Control 1.0 passed without a single MEP changing their mind. The vote that killed it in March and the vote that revived it in July had almost the same numbers, the difference was a procedural rule nobody outside Brussels was watching. Run the actual mechanics, because the “passed by default” framing is doing real work here. In March, the European Parliament voted 311 against, 228 in favor, rejecting an extension of the temporary law that lets platforms like Gmail, Snapchat, and Facebook Messenger voluntarily scan for known CSAM. That should have ended it. Instead, the EU Council adopted the Commission’s original text as a second-reading position in July, which triggered a different legal threshold entirely: Parliament could now only block it with an absolute majority of all 720 members, 361 votes, not simply more votes against than for. In the actual July vote, 314 MEPs opposed it, more than opposed it in March, and it still passed, because absences and abstentions count as support under that rule. Here’s what the amendment actually changed, and it matters more than the “chat control” branding suggests. MEPs explicitly exempted end-to-end encrypted services like WhatsApp and Signal from the law’s scope. This wasn’t a new concession; true end-to-end encryption already made server-side scanning technically impossible, since neither the platform nor anyone but the two parties involved holds the decryption key. What the exemption formally rules out is any future attempt to extend this specific temporary law into encrypted spaces, not a rollback of protections that were previously in place there. The law that did pass applies only to platforms that can already access message content directly, Gmail, Snapchat, Messenger, Skype, Xbox, using hash-matching against known CSAM databases, AI classification for new material, and text analysis for grooming patterns. It authorizes this scanning; it does not require it. Supporters, including the rapporteur who backed the extension, argue the alternative is a legal gap that stops platforms from continuing detection work that has directly led to arrests and rescues. Critics, including privacy researchers and MEPs like Markéta Gregorová, argue the procedural route used to revive a rejected bill undermines Parliament’s own rules regardless of the underlying merits of CSAM detection itself. The permanent framework, Chat Control 2.0, is still being negotiated, and it’s the version that would introduce mandatory client-side scanning even on encrypted platforms, the change privacy advocates consider the actual red line. This vote didn’t decide that fight. It just kept the temporary, voluntary version alive while it continues.

• 
  Chaitanya Pinapaka (@IamChaitu_) reported 48 minutes ago

  @gmail I am still getting this same error.

• 
  ☯️ SAFE LEAF/TREE BIRD (@chercher_ai) reported 55 minutes ago

  @Duderichy they're tired of people hounding them to fix Gmail and Google Docs