Microsoft's record set of security vulnerabilities. LastPass users warned, again.
Instagram image generation feature killed. And how the World Cup illustrates the changing face of online scams.
Patch pressure
This week saw Microsoft release updates for the biggest set of security issues in its history,underlining the challenge facing IT teams that we covered a couple of weeks ago. Two of the more than 600 vulnerabilities were already being used in real attacks before a fix existed, and dozens more were serious enough to allow attackers to take control of an affected system. Microsoft says this is the new normal: its engineers are now using AI to comb through the entire Windows codebase for flaws, and unsurprisingly, it’s finding a lot more of them than a team of humans ever could. And the trend isn’t limited to Microsoft. Adobe, Cisco, Mozilla, and Oracle are all increasing the speed at which they release updates, with Adobe explicitly citing AI as a factor.
Of course, when so many issues are rated “critical” in a single release, the label stops meaning much. As Tenable says, the systems used to rank how dangerous a flaw is were built with human attackers in mind, not AI models which can now work out what a patch fixes and build a working attack for it faster than most companies can finish testing. For IT teams, that shifts the ground rules. Increasingly, it means applying updates for any issues already being actively exploited, and prioritising systems like remote access tools, emails, and shared storage that can be reached from the outside world. The lower-risk issues have to go to the back of the queue. It may not be ideal but fixing more than 600 issues at once is obviously not practical. The choice isn’t between fast and careful any more. It’s between fast and too late.
Threats
Meta: Criminals have found a way to abuse Meta’s Business Manager tool to send phishing emails that genuinely come from Meta’s own email address, making them far more convincing than the usual fake notifications. Meta has now blocked the technique which was designed to capture a range of information, including MFA codes and even the target’s government issued ID. Huntress
Passwords: LastPass is warning about yet another phishing campaign targeting its users. The latest appears to notify recipients about updated security policies and directs them to a landing page that impersonates DocuSign and claims to provide a document for review. The emails are sent from hello@lastpassnewsletter.com. LastPass insists its own systems have not been compromised. Bleeping Computer
Microsoft: We’ve warned about “device code” phishing before because it’s proving remarkably effective. This works by hijacking a sign-in process meant for devices without a keyboard like smart TVs. Instead of stealing your password, attackers ask you to enter a legitimate login code on a genuine Microsoft page, thus authorising their device to access your account. ReliaQuest has a report showing how dedicated toolkits and AI are being used to automate these attacks, making them faster to launch and much harder to detect.
Apple: A particularly sneaky information stealer disguises itself as one of Apple’s own system tools, making it look genuine. Once opened, it displays a convincing password prompt which, if entered, gives criminals access to passwords stored in a Mac’s Keychain as well as browser data and cryptocurrency wallets. Apple doesn’t ask you to download system tools, so unexpected “Apple” apps should always raise suspicion. Jamf
AI
Credentials: Giving AI agents access to passwords has always sounded like a terrible idea. 1Password thinks it has found a better approach. A new integration with Claude allows the AI to use login credentials to complete tasks such as signing into websites, without ever seeing the passwords themselves. Each request still requires the user’s approval. We think this is likely the first of what will become a common integration. Though we can’t help feeling slightly uneasy about it.
Killed: Well, that didn’t last long. In fact, it took just three days for Meta to perform a swift about turn and withdraw the Instagram image generation feature that we wrote about last week. Not altogether surprisingly, there was an immediate backlash to a feature that allowed anyone to grab an image from a public account and do with it whatever they wanted - without notifying the owner. Mark Zuckerberg made a brief effort to style it out, saying safety measures were built into the tool, before the feature was taken behind the back of Meta HQ and humanely extinguished. “We’ve heard the feedback that this feature missed the mark, so it’s no longer available,” Meta said. At the time of writing, Meta’s website says it’s still available.
Alert: Meanwhile, Meta says it will now notify parents if their teenage child discusses suicide or self-harm with the company’s Meta AI chatbot. The feature is live for parents using Instagram Parental Supervision in the US, UK, Australia, and Canada, and will roll out globally by the end of the year. Meta also says it’s working on the ability to contact emergency services if someone’s conversations with its chatbot suggest they may be at risk of self-harm.
Censor: AI models might be restricting free expression, according to research by The Oversight Board, the independent content moderation organisation created by Meta. The group is calling for greater transparency about how AI companies decide what their systems will and won’t discuss, arguing that safety measures should not come at the expense of legitimate debate. Engadget
In brief
Crime: As well as rebranding advertising as hydration breaks, the World Cup has demonstrated how online scams are changing. The mass of World Cup-related scams showed how AI is being used to make fakes more credible and harder to spot. Scammers are also exploiting AI capabilities to scale up their campaigns. Instead of building a single site or message, AI allows them to generate multiple versions and test which works best. It means both technical tools and user awareness are vital to protecting individuals and organisations.
Morocco: The government used Pegasus spyware to surveil its citizens including a journalist, according to a former officer in the country’s internal intelligence agency. Morocco, which has long denied using Pegasus, is reported to have procured the spyware through an intermediary to avoid leaving a paper trail. OCCRP
Anonymity: Achieving and maintaining anonymity online is extremely difficult, as an alleged teenage hacker has just demonstrated. 19-year old Peter Stokes has been charged with computer intrusion, conspiracy and fraud. He also made the mistake several years ago of threatening a cybersecurity researcher who promptly notified her contacts about the incidents. That meant his activity had been tracked for years or, as she puts it, “He was a marked man before the events described in his criminal charges started. He got caught before he got careful.”
VPN: While on the subject of anonymity, the UK government has decided not to restrict the use of VPNs as part of its new social media controls. It said a survey of 11-17-year-olds had shown those who used the technology did so mostly for privacy, not to break the rules. Unhelpfully for the government, its decision coincided with research from the UK communications regulator that found use of VPNs in the country had almost doubled since the introduction of laws enforcing age limits on pornography sites last year.
Russia: The UK and EU are demanding urgent action from critical infrastructure organisations after formally blaming Russia’s Federal Security Service for a December 2025 cyberattack on Poland’s power grid. In particular, they highlight exploitation of poorly configured routers. Meanwhile, Truesec warns that Russian intelligence is also targeting home and small office routers in order to redirect traffic to phishing sites.
Offboarding: Apple claims a former engineer kept his work laptop and exploited a bug to access Apple’s cloud file storage while employed by OpenAI. The allegations have yet to be tested in court, and it’s also possible the access resulted from a failure to remove his network access rather than a software vulnerability. Either way, the case is a reminder that robust HR and IT offboarding processes are just as important as technical security controls.
Testing: As you probably know, USB-C cables aren’t all equal, and some aren’t what they claim to be. WhatCable is a free Mac app that reveals a cable’s true specifications - and could explain why some don’t do what you need them to do. ZDNet
Updates
Zoom: Emergency update for Windows Workplace app to address a critical vulnerability that could allow attackers to take over accounts remotely without user interaction.
F5: Emergency security updates to address eight vulnerabilities in NGINX and BIG-IP.
SonicWall: Customers are urged to immediately update SMA1000 secure remote access appliances, which are being targeted by attackers via two new zero-day vulnerabilities.