The race to keep up with vulnerability discovery. Targeting online shoppers.
Controlling AI costs. Work devices and the US Secret Service. Russia blamed for Jaguar Land Rover hack. And a new pixel that could turn displays into cameras.
Racing to keep up
Finding the time to keep solutions and devices updated has always been a challenge. Increasingly, the problem is finding it quickly enough. AI is accelerating the pace of vulnerability discovery - and exploits - to the point where companies are already replacing monthly release cycles with more frequent updates. Adobe has become the latest major software company to move from monthly to twice monthly releases, while Apple says AI-driven security concerns led it to bring forward its latest updates rather than waiting to include them in a new version of its iOS operating system. Some vendors are now beginning to talk about near real-time patching as the only way to keep up in the race between vulnerability disclosure and weaponisation.
The pressure isn’t theoretical. Last week, an anonymous researcher published exploits for more than a dozen previously unknown issues affecting widely used software including Linux, OpenVPN, VLC and 7-Zip, without first alerting their developers. The incident also highlights a broader shift. As AI makes vulnerability research faster and cheaper, organisations should increasingly assume that internet-facing services will come under pressure. That doesn’t mean abandoning VPNs or remote access, but it does mean reviewing every externally accessible system, ensuring the strongest possible authentication, and reducing the “attack surface” wherever practical. Given the pace of AI-driven vulnerability discovery, this may become a race that simply cannot be won. The priority therefore shifts from patching everything instantly to reducing the number of systems attackers can reach in the first place.
Threats
Shop: Shopify’s order-tracking app is increasingly being targeted by scammers who are adding fake purchase receipts to users’ order histories to trick them into providing sensitive data or installing remote access software. Online shopping is becoming a minefield of scams, and McAfee says many consumers tend to ignore red flags when searching for a good deal. Among their findings: 40% of people said they’d trust a lower-priced deal without verifying it.
AI: Attackers are creating fake ChatGPT/OpenAI sites that impersonate real companies. They then send legitimate-looking invites that pass email checks and may lure staff into sharing sensitive work in attacker-controlled projects. Push Security
Extensions: We’re bored of this. You’re bored of this. But we can’t ignore the extent to which browser add-ons/extensions are a risk. This week there’s an ad blocker for Chrome with malicious capabilities that lie dormant until activated. Microsoft shut down an operation targeting its Edge browser with a campaign of malicious extensions that had 2.6 million installs. And D3Lab explains how an extension is the key element in a multi-stage attack designed to compromise Chrome’s security settings.
Spray: An aggressive campaign targeting Microsoft 365 environments generated more than 81 million login attempts over a two-week period. The threat actor tried out still valid username and password combinations that had been exposed in past breaches. Which is why it’s worth adopting passkeys wherever you can. Huntress
Interpol: Smaller businesses are being targeted through phishing messages that appear to come from Interpol warning about current investigations. In fact, they deliver ransomware through a password-protected archive that purports to contain evidence. Bitdefender
AI
Expense: Organisations are throttling their employees’ use of AI and pleading with them to use less powerful models to stop AI costs from spiraling out of control, according to reporting by 404 Media. It analysed leaked Slack chats, screenshots of internal dashboards, emails, and other material from half a dozen companies including Atlassian, Adobe, and Amazon. In at least one case, monthly AI spending had tripled to more than $15 million.
Claude: Anthropic has released a restricted version of its Fable 5 model after the US government lifted export controls. The US Commerce Department said Anthropic had agreed to “proactively detect and address security risks associated with the models,” to work with the government on protocols for future releases and to report any “malicious activity” it finds in among its models. Meanwhile, Austria is reported to be lobbying Brussels to host Anthropic in the EU.
Privacy: Proton, best known for its ProtonMailVPN and encrypted email service ProtonMail, has announced the second iteration of its ‘privacy-first’ AI chatbot. Lumo which was announced last year is designed to be a privacy-preserving alternative to popular AI models such as OpenAI’s ChatGPT and Anthropic’s Claude.
Doctor: In 2024, 2% of people in the US reported turning to AI for health information. That figure now stands at 61%, according to Salesforce’s Connected Health Consumer report.
Persuasion: Research found that AI systems were reliably more persuasive than expert humans, even when the humans chose their issues, did advance research, underwent hours of live, structured practice, and were incentivised with £1,000 cash bonuses. AI’s advantage persisted after experts received a coaching tool that let them practice against the AI that beat them, review their performance, and see what AI would have said at key moments.
US
Service: We have extensive experience in the reluctance of employees to use their work devices. Turns out so does the US Secret Service. A federal review found that agents routinely used personal cell phones to communicate with law enforcement and each other because their government-issued devices lacked the capabilities they needed to perform their missions. In most cases, employees just want to get their job done and they’ll use whatever device or app they think does that best. That’s why organisations should ask people whether the solutions in use meet their needs. Otherwise, employees will make their own arrangements…
Bounty: The State Department is offering up to $10 million for information that helps to identify or locate hackers targeting WhatsApp and Signal users. The bounty appears aimed at groups linked to Russia’s intelligence and military services.
Location: The US Supreme Court has placed limits on the use of geofence search warrants by law enforcement agencies. The technique allows investigators to request geolocation and other data about individuals who were near a crime scene. The ruling means that specific search warrants with a suspect’s name will now be required to gain access to the data.
In brief
JLR: The New York Times reports that Russian hackers were behind last year’s devastating incident at Jaguar Land Rover which is estimated to cost the British economy some £2 billion. It quotes people familiar with the investigation as saying that the methodology of the attack did not fit with initial claims of responsibility from a hacking collective. Instead, they say a group of Russians was responsible though it’s not clear whether were acting at the behest of the Kremlin or with its tacit assent.
Exposed: Phishing attacks have exposed employee data at 86% of US Fortune 100 companies over the past 12 months, with the technology, airline and automotive sectors hit the hardest. SpyCloud says 78% of organisations in its survey had experienced an increase in phishing volume over the past year.
Spyware: Security researchers say a member of the European politician had his phone hacked with Pegasus spyware while serving on an investigatory committee examining abuses of commercial surveillance tools. Citizen Lab doesn’t say who was responsible but notes similarities with former attacks that targeted Russian and Belarussian-speaking journalists.
Chile: A court has sentenced a former judge and an ex-general to five years in prison for illegally spying on a journalist who was investigating corruption in the Chilean army. The court found they had fraudulently ordered a wiretap on the journalist, claiming his phone number belonged to a Bolivian migrant who they accused of being a foreign agent. CPJ
Unhidden: Last week, we reported that Apple is making its Hide My Email feature less effective. Now, it turns out that for at least a year it hasn’t been working anyway. A researcher demonstrated to 404 Media that it could unmask a “hidden” address. Worse they say they told Apple about it over a year ago but nothing has been done to fix it.
Snow: The Register has a regular column that collects security incidents “in hopes we can all learn from others’ mistakes.” The latest is about a couple of “red teamers” who are paid to break into offices and networks to test security. In one case, they gained access by walking into a client’s maintenance office (in the winter) where they posed as new IT employees. One of them offered to help clear snow, while the other was allowed in “to begin setting up a laptop.” That access was exploited to install a device to give them remote access which they then used to uncover security issues in the client’s network. It underlines why no-one should trust what visitors say, even if their story appears to check out…and they offer to help clear snow.
Display: Researchers in Switzerland have invented a new type of pixel that could create displays that also work as cameras. The bi-drectional pixels have already prompted concerns about privacy - while giving a new meaning to “on display.”
Updates
WhatsApp: In a major change, Meta has started rolling outusernames to its messaging platform. It means that instead of relying on phone numbers as the primary identifier, users will increasingly interact through platform-managed usernames. The idea is to improve privacy but it creates an obvious opportunity for impersonation. Meta says it’s blocked names for well known personalities, but TechCrunch found plenty still available, including “indiamodi.” Perhaps not surprising that the Indian government (Prop. Narendra Modi) has told Meta to suspend the change.
Teams: Microsoft is adding a feature designed to control whether automated bots can access meetings. It will prompt administrators to let selected bots through on a case by case basis. The feature is designed to allow access for meeting assistants and note-taking apps while blocking those designed to harvest (i.e. scrape) content.
Apple: iOS and iPadOS 26.5.2 and macOS 26.5.2 include multiple security fixes, most focused on web technologies.
Chrome: A bumper update for Google’s web browser which addresses no less than 382 issues. It’s never been more important to ensure your browser is updated. The easiest thing to do is to get into the habit of closing it down at the end of every day.
Opera: A new feature for the browser is designed to block “ClickFix” attacks. Paste Protect scans a user’s clipboard for malicious commands and displays a warning if it spots something suspicious. The feature is enabled by default and doesn’t need to be configured.
Citrix: Security updates to address multiple flaws in NetScaler ADC (formerly Citrix ADC) and NetScaler Gateway (formerly Citrix Gateway).
Tails: Version 7.9.1 updates Tor Browser and Tor client as well as addressing a vulnerability in the Linux kernel.