U.S. jobs growth was much slower than previously reported, according to revised data released on Tuesday. From a report: The number of jobs created in the United States from April 2024 to March 2025 was revised down by 911,000 by the Bureau of Labor Statistics. That would roughly amount to 76,000 fewer jobs created each month of the year up until March. The revision draws fresh attention to the weakening U.S. labor market, which added an average of only 29,000 jobs in each of the three most recent months.

The August jobs report showed that the U.S. added only 22,000 jobs that month and also revised June’s job growth down to a loss of 13,000 jobs. Those datapoints have led economists and some policymakers to conclude that the U.S. labor market is now at a standstill. “The jobs engine that has been integral to U.S. economic growth defying expectations for the past four years is stalling,” Sarah House, a senior economist at Wells Fargo, said in a note on Friday.

Read more of this story at Slashdot.

Did you know you can customize Google to filter out garbage? Take these steps for better search results, including adding Lifehacker as a preferred source for tech news.

AI continues to take over more and more of our day-to-day activities: Anthropic recently announced a Chrome extension that allows Claude AI to see browser activity and run actions on behalf of users, while Perplexity’s Comet is an AI-powered browser that the company calls both a “personal assistant” and a “thinking partner.”

Agentic browsers may be able to do a lot of things for you, such as scheduling meetings, replying to emails, and ordering DoorDash, but handing all of this control (and personal information) over to AI comes with potential security risks. One of these is a prompt injection attack, which allows hackers to trick the AI into following their instructions instead of yours.

What is a prompt injection attack?

A prompt injection attack is when hackers disguise malicious inputs to AI as legitimate ones, so generative models are tricked into divulging sensitive data or taking harmful action.

As IBM describes, large-language models (LLMs) are given sets of instructions—system prompts—for how to handle user inputs. These two elements are combined into a single command, both written in natural language, which means that the LLM cannot separate which part of the command is the system prompt and which comes from the user. If threat actors create an input that bears enough resemblance to a system prompt, it could supersede the legitimate developer instructions and force the LLM to follow the fake ones.

In practice, this may involve hiding malicious prompts on a webpage the LLM is likely to read in order to carry out an action. The content, which could be plain text or embedded in an image or PDF, may look harmless or be invisible to users (employing white text on a white background, for example). Hackers don’t need code to carry out a prompt injection attack—just the right words in the right place.

How prompt injection compromises agentic browsers

While browsers with AI integration still require some manual input to complete tasks, agentic browsers act more like autonomous assistants that can follow entire workflows without user approval. That means that there’s no safeguard of human review before AI potentially shares your information, runs a malicious program, or spends money on a fraudulent purchase.

An example from Malwarebytes Labs: You ask your agentic browser to find and book a cheap flight for your next vacation. If it has all of your passenger and payment information available (because you’ve provided it), AI can complete this request without any additional action from you. But if the cheapest flight is found on a malicious website set up for this purpose, the browser could hand your credit card number and other sensitive data directly to the scammers.

A recent report from researchers at Brave (which has its own AI assistant) outlines particular concern about Perplexity’s Comet, with tests showing that the agentic browser is vulnerable to prompt injection attacks and hasn’t yet fixed the issue. Anthropic, for its part, has acknowledged its vulnerabilities and notes that it is working on safeguards to minimize them.

How to safely use agentic browsers

Mitigating prompt injection attack risks falls largely on the developers of agentic browsers rather than the user, with security experts recommending higher standards for user interaction and distinguishing between a user’s request and other content consumed to carry out an task.

That said, while Perplexity and Anthropic and others address these issues on their end, you can put guardrails in place against prompt injection, such as limiting the data and accounts your agentic browser can access and requiring manual review for high-stakes tasks, such as authorizing payments. Malwarebytes Labs also recommends enabling multi-factor authentication on all accounts connected to agentic browsers, regularly reviewing account and browser activity, and keeping software updated to ensure security flaws are patched in a timely manner.

Nintendo-watchers may remember when little-known accessory-maker Genki showed the world an extremely accurate 3D-printed mockup of the Switch 2 back in January, about a week before Nintendo’s own “first look” teaser video and months before the system’s wider unveiling. Now, in a newly filed settlement agreement, Genki has agreed to pay Nintendo unspecified damages for “trademark infringement, unfair competition, and false advertising” in connection with that promotional stunt.

The controversy started at the Consumer Electronics Show, where Genki showed off to press and attendees the 3D mockup of the then-unseen Switch 2 sporting the company’s accessories. Genki also posted a video showing a detailed mockup of the system from various angles. By April, Genki was promoting a “Genki Direct” livestream focused on “upcoming accessories for Genki Glitch 2” (no that isn’t a typo).

Nintendo’s lawyers reportedly visited Genki’s CES booth once reports of the Switch 2 mockup started going viral. But a Genki representative asserted at the time that the company did not sign an NDA with Nintendo and therefore “has nothing to worry about.”

Read full article

Comments

Did you know you can customize Google to filter out garbage? Take these steps for better search results, including adding Lifehacker as a preferred source for tech news.

Signal users will soon be able to save end-to-end encrypted backups of their chat history in a secure archive, making it possible to restore your conversations even if you lose your device. While text message and limited media backups will be free, the platform is also offering longer-term media storage as its first paid feature.

Secure backups are available now to users running the latest Android beta, with a full public release to those on iOS and desktop coming “soon.” You’ll see the opt-in for secure backups in your Signal Settings menu upon rollout.

Signal’s backup plans

With secure backups, anyone will be able to opt into full text message backups (up to 100 MiB) and media backups—including photos, videos, GIFs, files, and attachments—from the previous 45 days at no cost. Or you can pay $1.99 per month for full media backups beyond 45 days with a storage limit up to 100 GB.

If you enable secure backups, your device will automatically create a new archive daily and replace the one from the day before. View-once messages and those set to disappear in the next 24 hours are excluded.

How Signal keeps your backups secure

Signal is an open-source messaging platform with a reputation for strong privacy and security settings and features, such as end-to-end encryption and disappearing messages, so you can reasonably expect it to take protecting your backups seriously. According to the blog post outlining the new feature, backup archives will be stored without any direct link to your Signal user account or subscription payment, so they cannot be tied back to you.

Backups are secured behind a 64-character recovery key generated on your device, meaning it is never shared or stored on Signal’s servers. (This is different from your Signal PIN.) Your recovery key is the only way to unlock your backups, so if you lose it, you’ll permanently lose your archives too. Make sure you store this key in a secure spot, such as a password manager.

Other related features coming in the future include the option to save your backup archive to a location of your choice and the ability to transfer your encrypted message history between devices.

The Anker Soundcore Space A40 wireless earbuds are back on sale for a record-low price of $45 a 44 percent discount. These earbuds remain our top pick for best budget wireless earbuds.

The most significant feature on the A40 earbuds is the inclusion of active noise cancellation (ANC), which is almost unheard of at this price point. It’s not the most sophisticated ANC, but it does sufficiently reduce constant rumbling background noise like on a plane. The built-in mic that’s used for transparency mode and phone calls is a bit subpar, especially when used in noisy areas.

These are budget-friendly, so while they can’t be expected to compete on audio quality with flagship headphones that cost hundreds of dollars, the sound is still impressive given their low cost. The sound profile is warm, with a strong upper-bass, while some of the higher frequencies lose finer details. For those wanting to fine-tune the sound on the A40s, the earbuds can be paired with the Soundcore app, which includes an EQ.

The small, round earpieces help make the A40s comfortable to wear for extended periods, and the charging case extends the already solid 10-hour battery life by another 50 or so hours. The Anker Soundcore Space A40 wireless earbuds are already one of the best value options on the market, and this sale makes them even more appealing.

Follow @EngadgetDeals on X for the latest tech deals and buying advice.

This article originally appeared on Engadget at https://www.engadget.com/deals/our-favorite-cheap-wireless-earbuds-are-back-on-sale-for-only-45-152704765.html?src=rss

The Plex streaming platform has experienced a security breach and is telling customers to change their passwords “immediately.” They also suggest that users enable two-factor authentication and sign out of any connected devices that are currently logged in.

The company says a database was accessed by an “an unauthorized third party” and that some customers had their emails, usernames and hashed passwords exposed. As indicated, the breach involved hashed passwords, which are scrambled through an algorithm, so it’s unlikely the bad actor could read them. However, this is an absolute case of “better safe than sorry.”

Plex says credit card data is safe, which is a relief, as that information isn’t stored on its servers. The company says it is currently safeguarding its data and “undergoing additional reviews to ensure that the security of all of our systems is further strengthened to prevent future attacks.”

This whole ordeal is eerily reminiscent of a Plex data breach in 2022, in which a bad actor infiltrated the system and accessed people’s emails, usernames and passwords. So consider this your reminder to change passwords at regular intervals and to consider using a dedicated password manager. Authenticator apps can also be helpful. It’s the wild west out there.

This article originally appeared on Engadget at https://www.engadget.com/big-tech/plex-tells-users-to-change-their-passwords-after-data-breach-152635869.html?src=rss

Apollo Chief Economist Torsten Slok, writing in a blog post There are fewer public companies to invest in, and firms that decide to do an IPO are getting older and older.

In 1999, the median age of IPOs was five years. In 2022, it was eight years, and today, the median age of IPOs has increased to 14 years.

The rise in the age of companies going public is not only a result of the Fed raising interest rates in 2022, but also the consequence of more companies wanting to stay private for longer to avoid the burdens of being public.

Combined with the domination of passive investing, failure of active managers and high correlation in public markets, and high concentration in a few stocks, the reality is that there is no alpha left in public markets.

Read more of this story at Slashdot.

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

Did you know you can customize Google to filter out garbage? Take these steps for better search results, including adding Lifehacker as a preferred source for tech news.

For a pair of earbuds that first made waves back in 2022, the Anker Soundcore Space A40 still manages to feel current—and the price drop to $49.89 (down from $99.99), which is the lowest they’ve ever gone (according to price trackers), makes them worth a second look. Back when they launched, PCMag gave them an Editor’s Choice and named them the best budget earphones of the year. The fundamentals for that haven’t changed much. You still get balanced audio that’s surprisingly detailed for the price, along with active noise cancellation that holds its own against more expensive models.

Soundcore by Anker Space A40 Auto-Adjustable Active Noise Cancelling Wireless Earbuds

$49.89
at Amazon

$99.99
Save $50.10

$49.89
at Amazon

$99.99
Save $50.10

In terms of design, the A40s are compact and comfortable, with silicone tips in multiple sizes to help you get a good seal. Inside, 10mm drivers deliver a frequency range of 20Hz to 40kHz, and the sound leans balanced rather than bass-heavy, though you can tweak it in the app if you want more punch. Android users get the bonus of LDAC support, which is usually reserved for pricier models, but even iPhone owners will be fine sticking with AAC. The app also lets you customize the controls, since the default double-tap gestures feel a little limited, notes this PCMag review.

The noise cancellation of these IPX4-rated earbuds does a good job with the constant low rumble of planes or traffic, though, reportedly, sharper, busier environments still get through a bit. You might also notice the sound shifting slightly when ANC is on, but it’s nothing you can’t adjust for in the EQ. Battery life is around 10 hours per charge with ANC off, and up to 50 more from the case. Even if those numbers shift in real-world use, they’ll easily last through a day or two of heavy listening. At the end of the day, for under 50 bucks, the A40s cover the basics well and even punch above their weight in a few areas.

AI summaries seem to be everywhere, with the latest entry into the mix coming courtesy of Mozilla. The company has announced that its web browser, Firefox, is rolling out page summaries on mobile devices. Notably, this feature is only available for English-language iOS users in the US right now.

Unlike other major summary tools from Google and the like, Firefox’s page summaries are available when you shake your device. If you feel a bit weird doing that then you can, instead, tap the thunderbolt icon on the right side of the address bar. There’s also the option to click the three-dot menu icon and then tap summarize page. 

You should be able to get a summary of any webpage with 5,000 words or less — though, that might require a lot of guesstimating if the word count is anywhere close. You can also turn off summaries in settings.

Firefox will use Apple Intelligence for the summary if you have an iPhone 15 or newer once iOS 26 drops. Otherwise, Mozilla’s cloud-based AI should securely receive the page text and then create a summary. 

The latest summary feature follows Firefox’s Link Previews, launched in May. The experimental feature used AI to preview a link’s content, including a list of key points. 

This article originally appeared on Engadget at https://www.engadget.com/ai/firefox-provides-ai-page-summaries-if-you-shake-your-iphone-145837557.html?src=rss

An Associated Press investigation based on tens of thousands of leaked documents revealed Tuesday that American technology companies designed and built core components of China’s surveillance apparatus over the past 25 years, selling billions of dollars in equipment to Chinese police and government agencies despite warnings about human rights abuses.

IBM partnered with Chinese defense contractor Huadi in 2009 to develop predictive policing systems for the “Golden Shield” project, AP reports, citing classified government blueprints. The technology enabled mass detentions in Xinjiang, where administrators assigned 100-point risk scores to Uyghurs with deductions for growing beards or being aged 15-55. Dell promoted a laptop with “all-race recognition” capabilities on its WeChat account in 2019. Thermo Fisher Scientific marketed DNA kits as “designed” for ethnic minorities including Uyghurs and Tibetans until August 2024.

Oracle, Microsoft, HP, Cisco, Intel, NVIDIA, and VMware sold geographic mapping software, facial recognition systems, and cloud infrastructure to Chinese police through the 2010s. The surveillance network tracks “key persons” whose movements are restricted and monitored, with one estimate suggesting 55,000 to 110,000 people were placed under residential surveillance in the past decade. China now has more surveillance cameras than the rest of the world combined.

Read more of this story at Slashdot.

Anthropic has begun rolling out a small but significant update to Claude. Starting today you can use the chatbot to create and edit Excel spreadsheets, documents, PowerPoint slide decks and PDFs. In the past, Claude offered rudimentary file support, but now you can interact with any documents you need to modify directly through the chatbot. The new functionality is part of a feature preview you can try out as long as you have a Max, Team or Education subscription. Sorry, Pro and free users, you’ll have to wait. The preview will roll out to Pro subscribers “in the coming weeks,” with no timeline yet for when free users can expect access.

“We’ve given Claude access to a private computer environment where it can write code and run programs to produce the files and analyses you need. This transforms Claude from an advisor into an active collaborator. You bring the context and strategy; Claude handles the technical implementation behind the scenes,” says Anthropic of how it built the feature. “This shows where we’re headed: making sophisticated multi-step work accessible through conversation. As these capabilities expand, the gap between idea and execution will keep shrinking.”

To check out the preview, toggle “Upgraded file creation and analysis” in the settings menu, which you can find by first selecting “Features” and then “Experimental.” You can then upload or describe the file you’d like Claude to create or edit for you, and download Claude’s creation once you’re happy with the result. 

This article originally appeared on Engadget at https://www.engadget.com/ai/claude-can-now-edit-and-create-files-including-excel-spreadsheets-143043937.html?src=rss

It has often been said that the lawyer who represents himself has a fool for a client. That truism apparently extends to Switch modders as well, at least in the case of Ryan Daly. The operator of the now-defunct moddedhardware.com (archived) now owes Nintendo $2 million after a recent settlement filed in a Washington state District Court.

The final judgement, which was jointly agreed to by both parties, asserts that Daly knowingly sold devices such as the MIG Switch and MIG Dumper that “are primarily designed for the purpose of circumventing the [technological protection measures, and] have limited commercially significant purpose or use other than to circumvent the TPMs.” Daly also sold and installed mod chips to get around Switch DRM and “copied and distributed certain copyrighted Nintendo games to his customers,” according to the judgment.

Nintendo’s lawsuit alleged that the company initially approached Daly about the sale of these devices in March 2024, and that Daly agreed at the time “both verbally and in signed writing” to refrain from further sales. After Daly continued to sell the devices, though, Nintendo took the further step of bringing this lawsuit last June.

Read full article

Comments