Only 15% of workers in Singapore feel safe from job cuts

Singapore workers rank among the least confident in the world about their job security—and the numbers are stark. Just 15% of employees feel their role is safe from elimination, according to ADP Research’s latest People at Work report.

Only four other markets rank lower than Singapore. In Europe, only 12% of workers in the Czech Republic feel secure in their jobs, while the remaining three are all in APAC—Taiwan (11%), South Korea (9%), and Japan, dead last at just 5%.

While Singapore ranked low, the broader region fared only slightly better, with just 18% of respondents across APAC saying they believed their jobs were safe from elimination. Globally, the figure stood at 22%.

This comes despite worldwide unemployment remaining at its lowest level in decades.

Jessica Zhang, Senior Vice President of APAC at ADP, said the results point to a disconnect between current labour market conditions and employees’ perceptions of their long-term job security.

“In Singapore, employees are not thinking about whether they have a job today, but also whether their roles will remain relevant tomorrow. This growing uncertainty is becoming a defining feature of today’s workforce,” she said in a statement.

The executive stressed the importance of clearer communication and continued upskilling efforts as concerns over job security grow.

Zhang said employers should be more transparent about how roles are changing, what those shifts mean for employees, and invest in ongoing training to help workers stay employable while boosting productivity and efficiency.

Long hours, low engagement

Singapore workers also reported high levels of unpaid overtime.

Overall, 45% of workers in Singapore said they clocked more than five unpaid hours a week, with 35% saying they put in between six and 15 unpaid hours each week, while another 10% said they worked 16 or more unpaid hours weekly.

The proportion of workers doing six to 15 unpaid hours exceeded the APAC average of 30%, while the share working 16 or more unpaid hours was in line with the regional average of 10%.

At the same time, Singapore recorded relatively high levels of generative AI adoption. Around 23% of workers said they used AI almost every day, while just 8% said they had never tried the technology.

Despite this, employee engagement levels in Singapore remained relatively low. Fully engaged workers accounted for 12% of the workforce in both 2024 and 2025, below the APAC average of 15%.

ADP Research’s People at Work report gathered responses from more than 39,000 adult workers in 36 markets between Jul 21 and Aug 4, 2025.

In APAC, the survey covered 13,136 respondents across multiple markets, including Singapore, Australia, China, India, Indonesia, Japan, New Zealand, the Philippines, South Korea, Taiwan, Thailand, and Vietnam.

• Read the full report here.
• Read other articles we’ve written on Singapore’s current affairs here.

Featured Image Credit: TK Kurikawa/ Shutterstock.com

Security

‘Budgets are moral documents,’ Rep. Delia Ramirez said

Democratic lawmakers on Thursday blasted President Trump’s spending priorities – specifically a proposed $1 billion White House security and ballroom project and a nearly $1.8 billion “slush fund” for Trump allies tied to the January 6 Capitol riot – as his administration pushes deep cuts to cybersecurity funding.

US Representative Delia Ramirez (D-IL) decried the president’s priorities as Congress weighs reauthorization of the State and Local Cybersecurity Grant Program (SLCGP), a funding effort that began in 2022 and earmarked $1 billion to state and local governments over the next four years to help mitigate cyber risks. 

“Budgets are moral documents, and spending a billion dollars on a ballroom, which is what the president wants, or $1.7 billion to incentivize insurrectionists while we still are waiting for the reauthorization of this critical grant program, says a lot about where priorities are right now with this administration,” she said during a House Homeland Security subcommittee hearing on state and local cybersecurity.

Another Democrat on the committee, Rep. James Walkinshaw (D-VA), noted the US Cybersecurity and Infrastructure Security Agency (CISA) also eliminated federal support for the Multi-State Information Sharing and Analysis Center (MS-ISAC), which used to provide free and low-cost threat detection and response services to state and local governments. 

The MS-ISAC has since shifted to a fee-based model to support the state threat sharing program. 

This means, as expert witness Samir Jain, VP of policy for the Center for Democracy and Technology, testified, “jurisdictions that most need the help are least likely to be able to afford it. Smaller jurisdictions, because if they don’t have the resources and the money to join the ISAC, they probably also don’t have the resources and the money to buy equipment, to buy network monitoring tools, to have cybersecurity staff. It’s the ones who need it the most are the least likely to be able to get it as a result.” 

Walkinshaw also pointed out that CISA’s 2025 budget was about $3 billion. President Trump proposed slashing the cyber-defense agency’s spending by $707 million in 2027, to just over $2 billion. 

This is on top of the $135 million in cuts to CISA, along with about a third of its workforce (close to 1,000 people) since Trump returned to office. 

“So we are looking at a one-third cut in federal funding for cybersecurity,” Walkinshaw said. “If President Trump gets his way, we’d be spending a billion dollars for the ballroom and $1.8 billion for the January 6 slush fund – $2.8 billion just on those two items, $800 million more than his total commitment to cybersecurity.”

Meanwhile, other expert witnesses who testified before the committee, all IT and security chiefs from Tennessee, New York, and Florida, implored the lawmakers to spend more – not less – on state and local infosec.

“State and local governments operate critical systems that citizens rely on every day, including emergency services, schools, utilities, courts, and public infrastructure,” Tennessee CIO Kristin Darby told lawmakers. 

“Those systems are increasingly targeted by criminal organizations and nation-state actors,” she said, adding that “demand for cybersecurity support far exceeds the current funding levels.”

As AI-enabled attacks, ransomware infections, and cloud-based system intrusions accelerate across Tennessee, “many local governments across our state have little or no dedicated cybersecurity staff,” Darby continued. “This creates a dangerous imbalance between highly sophisticated attackers and severely resource-constrained defenders.”

New York state director of security and intelligence Colin Ahern urged lawmakers to “reauthorize and fully fund the state and local cybersecurity grant program, which is the single most consequential investment in the cyber protection of state and local governments in this country.”

He also advocated for frontier-model AI access for state and local governments, which are tasked with protecting the power grid, drinking water supply, public health systems, and other critical operations. 

“We cannot do that while frontier defensive AI capabilities are restricted to federal partners and a handful of large enterprises,” Ahern said. 

“Cybersecurity is the silent partner of democracy,” he continued. “When the utilities, school districts, and state and local governments that constitute the operational fabric of American life are hollowed out by cyber attacks, the institutions that support our democratic life are hollowed out with them.” ®

Xiaomi CEO Lei Jun had a straightforward reason for why the YU7 wasn’t outselling the Tesla Model Y in China: the base model wasn’t cheap enough. 

At just 10,000 yuan ($1,450) less than Tesla’s Model Y, the price gap simply wasn’t compelling enough. On the evening of May 21, at Xiaomi’s “Human x Car x Home” launch event, Lei did something about it (via CarNewsChina).

Introducing the Xiaomi YU7 Standard Edition

The new True Standard Edition is priced at 233,500 yuan, about  $34,300, which makes it 30,000 yuan ($4,350) cheaper than the standard rear-wheel-drive Model Y. 

While that price gap should be enough, the more interesting number is the range figure. The YU7 lasts up to 399 miles on a single charge, compared to 368 miles for the equivalent Model Y variant. 

To put it simply, Xiaomi is offering more range for substantially less money, which, in China’s super-competitive EV market, is as direct a challenge as a company can issue.

What changed under the hood?

The YU7 True Standard Edition sports a rear-wheel-drive single-motor setup that produces 235kW, with a CATL-supplied lithium iron phosphate battery. 

Dimensionally it maintains the YU7’s mid-to-large SUV proportions at five meters long. However, it comes in 253 pounds lighter than the previous standard version at 4,850 pounds. 

As for why Xiaomi is doing this right now, the company wants to regain lost sales momentum. The YU7 arrived on June 18, 2025 at prices below $48,500 and secured over 200,000 orders within three minutes, creating a waitlist that stretched nearly a year. 

That backlog has since cleared, and with it, the sales momentum. The company sold fewer than 10,000 units last month, a significant drop from the launch frenzy, and, as a result, the company has launched the new Standard Edition. 

Most users can start generating slides with AI today.

The AI industry has fully entered the “agent era,” a paradigm where AI models do far more than generate text — they now actively plan, execute, and course-correct complex tasks over days rather than seconds.

Thus, it’s perhaps unsurprising to see Chinese e-commerce giant Alibaba’s famed Qwen Team of AI researchers release a model capable of performing autonomous agentic AI work over multiple days: that model has arrived in the form of Qwen3.7-Max which the company reports in a blog post achieved “~35 hours of continuous autonomous execution” — albeit, in a proprietary, not open source format, as prior Qwen Team releases were.

This is also to be expected — it’s what many analysts and industry experts feared in the wake of the departure of several key Qwen Team leaders earlier this year. But it makes sense for Alibaba financially, at least in the short term: training AI models, especially ones as powerful as Qwen3.7-Max, is expensive, and giving them away essentially for free, as open source models are, does not immediately help recoup any costs.

In that sense, Alibaba is simply aligning its efforts with American AI giants like OpenAI and Google by offering the latest and greatest models only through paid APIs and subscription or paid web plan bundles, and slightly less performant ones through open source.

Still, the arrival of Qwen3.7-Max offers further optionality to enterprises and individual users, and more competition for American AI labs — rarely a bad thing for consumers at all budget levels. Yet, the fact that the model is only accessible from Chinese-based endpoints means it may be limited in its appeal to American and European enterprises seeking to maximize compliance and security posturing when fulfilling government contracts, or even just attempting to comply with all relevant state, local, and national data sovereignty regulations.

The marathon AI era

To understand why Qwen3.7-Max is a departure from previous models, one must look at how it was trained and how it operates in practice.

Language models typically degrade when forced to maintain a single train of thought over thousands of conversational turns; they forget instructions, hallucinate variables, or simply get stuck in logical loops. Qwen3.7-Max was specifically designed as a “versatile agent foundation” capable of “long-horizon reasoning” to overcome this exact bottleneck.

The starkest demonstration of this capability is an autonomous engineering task detailed by the Qwen team. The model was given access to an isolated server equipped with a T-Head ZW-M890 PPU—a hardware architecture the model had never encountered during its training. Its task was to optimize an attention kernel.

Over the course of 35 straight hours, Qwen3.7-Max operated entirely autonomously. It executed 1,158 distinct tool calls, performed 432 kernel evaluations, diagnosed compilation failures, and iteratively improved the code to achieve a 10.0x geometric mean speedup.

By comparison, Chinese competitor models like z.ai’s GLM-5.1 and Moonshot’s Kimi K2.6 capped out at 7.3x and 5.0x speedups respectively, often voluntarily terminating their sessions when they failed to make progress. However, both are available open source.

This endurance is achieved through what Alibaba calls “environment scaling”. Just as early LLMs grew smarter by ingesting more diverse text, Qwen3.7-Max was trained across a vast, scaled array of dynamic agentic environments.

It is capable of simulating a one-year lifecycle of a startup in the “YC-Bench” evaluation, navigating hundreds of decision-making rounds encompassing personnel management and contract screening. In this simulation, the model managed to generate $2.08 million in virtual revenue, nearly doubling the performance of the prior generation, Qwen3.6-Plus.

Furthermore, the model has built-in reward-hacking self-monitoring, autonomously detecting when it attempts to cheat a training environment and adding heuristic rules to correct its own behavior.

A brain for any scaffold

From a product perspective, Qwen3.7-Max is designed to be the cognitive engine for modern software development and enterprise automation.

The model offers a massive 1-million-token context window and a 64K maximum output limit, providing immense overhead for processing sprawling codebases or lengthy technical documents.

One of its most compelling features is “cross-harness generalization”. Rather than being hardcoded to work best within a specific proprietary interface, Qwen3.7-Max is built to act as a drop-in intelligence layer for diverse agent frameworks. It supports the Anthropic API protocol natively, allowing developers to plug it directly into existing tools like Claude Code or OpenClaw.

The benchmark data provided by Alibaba indicates that this generalized approach has paid massive dividends.

On the Apex Math Reasoning benchmark, Qwen3.7-Max scored 44.5, eclipsing Claude Opus-4.6 Max’s score of 34.5 and DeepSeek V4-Pro Max’s 38.3. It also posted dominant scores on Humanity’s Last Exam (41.4) and the realistic coding agent benchmark MCP-Atlas (76.4).

This translates into tangible utility for end-users. Through open source Model Context Protocol (MCP) integrations, the model can operate as an autonomous office assistant, capable of reading university formatting specs and automatically reformatting a messy Word document via command-line tools without human intervention.

Running this level of intelligence comes at a distinct cost. Developers accessing the API via Alibaba Cloud Model Studio will pay $2.50 per 1 million input tokens and $7.50 per 1 million output tokens. The platform also features explicit cache creation and read pricing, as well as a $10 fee per 1,000 calls for integrated web searches, though code interpreter tools remain free for a limited time.

Qwen3.7-Max occupies a strategic middle ground in the current API economy. While it demands a notable premium over aggressively priced domestic rivals—costing nearly double DeepSeek V4 Pro ($5.22) and Z.ai’s GLM-5.1 ($5.80)—it drastically undercuts the Western frontier giants it routinely matches on benchmarks.

For context, running heavy agentic workflows through OpenAI’s GPT-5.4 or Anthropic’s Claude Opus 4.7 will run developers $17.50 and $30.00 per million tokens, respectively. See VentureBeat’s pricing chart below:

By positioning Qwen3.7-Max just below Google’s Gemini 3.5 Flash ($10.50) but well above budget-tier models, Alibaba is signaling that this isn’t a commodity release; it’s a flagship reasoning engine priced to lure enterprise workloads away from Silicon Valley’s most expensive offerings.

Licensing remains proprietary for now

For all its technical brilliance, the most controversial aspect of Qwen3.7-Max is how it is distributed. Qwen is billing the release as a “proprietary model”. It is strictly API-only.

Historically, Alibaba’s Qwen has been a hero to the open-source and local LLM communities. Previous iterations, like Qwen 2.5 and Qwen 3.6, released their weights publicly. Open weights allow developers, researchers, and enterprises to download the model, run it on their own hardware, and fine-tune it for highly specific or data-sensitive use cases without sending proprietary information to a third-party server.

By locking Qwen3.7-Max behind an API, Alibaba is pivoting to the standard commercial playbook utilized by OpenAI (with GPT-4) and Anthropic (with Claude). For enterprise users, this means utilizing Qwen3.7-Max requires trusting Alibaba Cloud with their data streams and relying entirely on internet connectivity to run their agentic workflows. For the open-source community, it means losing access to what is currently one of the most capable models on the planet.

Community reactions split between awe and disappointment

The reaction from the developer community has been swift, characterized by a mix of profound respect for the engineering achievement and frustration over the licensing model.

Prominent AI commentator Sudo su (@sudoingX) captured the prevailing sentiment on X (formerly Twitter). “qwen is unreal,” they wrote. “they just dropped 3.7 max and it is beating opus 4.6 max on most of the benchmarks they ran”.

The technical metrics, particularly the model’s endurance, have left many in the field stunned. “the apex math number, 44.5 against opus 34.5, that is not a small gap,” Sudo su noted. “the 35 hours straight on a kernel optimization task with 1000+ tool calls is the part i keep rereading. that is the agent era thing actually happening, not a slide”.

The speed of Alibaba’s iteration is also drawing notice. With Qwen 3.6 released just last month, the leap to 3.7-Max highlights a relentless development cadence. As Sudo su observed, “nobody else is moving like this”.

Yet, the praise is heavily caveated by the shift to a closed ecosystem. The loss of the model weights is seen as a blow to the localized AI movement, which relies on state-of-the-art open models to push the boundaries of what can be done on consumer hardware or private enterprise clusters.

“one thing though, please open source this one too,” Sudo su pleaded in their post. “3.6 dense made the entire local llm ecosystem better. the max tier going api only would close a door we have been keeping open. give us the weights eventually”.

Qwen3.7-Max proves that the autonomous agent era is no longer a theoretical projection; it is a present reality capable of executing complex engineering feats while humans sleep. The only question now is whether this new frontier of AI will be a democratized resource you can download to your laptop, or an intelligence utility rented strictly from the cloud. For now, with Qwen3.7-Max, it is undeniably the latter.

from the ctrl-alt-speech dept

Ctrl-Alt-Speech is a weekly podcast about the latest news in online speech, from Mike Masnick and Everything in Moderation‘s Ben Whitelaw.

Subscribe now on Apple Podcasts, Overcast, Spotify, Pocket Casts, YouTube, or your podcast app of choice — or go straight to the RSS feed.

In this week’s roundup of the latest news in online speech, content moderation and internet regulation, In this week’s roundup of the latest news in online speech, content moderation and internet regulation, Mike is joined by civil liberties lawyer Jennifer Granick. Together they discuss:

Support the podcast by joining our Patreon, with special founder membership available until May 28th.

Filed Under: africa, ai, artificial intelligence, content moderation, deepfakes, trust and safety

Companies: apple, kickstarter, twitter

The Federal Trade Commission announced on Thursday that Cox Media Group and two other marketing companies, MindSift LLC and 1010 Digital Works, have agreed to collectively pay nearly $1 million to settle allegations that they deceived their customers—other businesses—by claiming that they could help target ads based on audio recordings collected from consumers’ smart devices via a marketing service called Active Listening.

In a statement to WIRED, a spokesperson for CMG says, “We are pleased to have this matter resolved. Our local marketing team relied on marketing materials provided to us by a third-party vendor about their product. We withdrew the materials expeditiously and stopped further use of the product.”

MindSift and 1010 Digital Works did not immediately respond to a request for comment. (Disclosure: The author of this article previously worked for the FTC.)

Over the years, conspiracy theories about companies listening to people through their phones in order to serve them ads have been repeatedly debunked. The marketing about Active Listening, which was first reported by 404 Media, stoked those fears. According to the FTC, at one point a website advertising the service included the slogan, “Creepy? Sure. Great for marketing? Definitely.”

In three separate complaints, the FTC says that CMG made several claims about its ability to collect consumers’ conversations from “smartphones, smart TVs, smart speakers and other devices” and then use AI to target ads to potential customers based on where they live and what they said. CMG and the other companies also said that consumers had consented to the collection and use of their voice data, according to the complaints.

The FTC alleges that none of those things were true.

Instead, the FTC contends that what CMG was offering was “nothing more than consumer email list buying” and that the lists it resold were “a significant markup over the cost of the data.”

As part of their agreements with the FTC, CMG and the two other companies promised not to make misrepresentations about their marketing services or their collection and use of audio recordings or transcripts of consumer conversations.

CMG agreed to pay $880,000, while MindSift and 1010 Digital Works each agreed to pay $25,000. The combined $930,000 will go to businesses that were “impacted” by the three companies’ practices, according to the FTC—in other words, businesses that purchased the Active Listening marketing service because they were under the impression that the service worked as advertised, including that people consented to having their voice data used.

The FTC’s complaints don’t make allegations about whether it’s illegal to use audio recordings collected from people’s smart devices to target them with ads, but the FTC clearly has a problem when a company says it does that but actually doesn’t. In a statement, Christopher Mufarrige, the FTC’s director of the bureau of consumer protection, says, “It is a basic rule of business that you need to be honest with your customers, and these companies failed to do that.”

The Supreme Court could now weigh in on the Apple versus Epic case where Apple was found in contempt of an injunction and forced to allow all developers to link externally without commission.

The Apple versus Epic saga is nowhere near an end even if Epic is celebrating a victory prematurely. Even as the case returns to Circuit Courts, Apple is requesting the Supreme Court to review two specific issues it has with the proceedings so far.

In the Supreme Court filing viewed by AppleInsider, Apple shares that the scope of the anti-steering injunction exceeds the District Court’s limits set by CASA. It also argues that the injunction violation was issued in error due to suggesting it was violating the “spirit” of the law rather than the letter.

Its arguments in the 34-page filing suggest that the Supreme Court should take up these matters because Apple’s is a perfect vehicle to address these issues. Apple asserts that providing a decision would settle matters for future cases, and if left untouched, could cause the CASA verdict to be a dead letter.

Basically, Apple hopes that there are enough discrepancies to ensure the Supreme Court at least picks up the case. In the meantime, Apple will continue its proceedings with Epic in the lower courts.

The story so far

Epic sued Apple in 2020 on antitrust grounds, but Epic lost on every count except one. That count pertained to Apple’s anti-steering practices.

Epic’s ‘1984’ parody ad

Apple removed the anti-steering provisions and provided a new, if complex, way for developers to link to external purchases. It meant developers still owed Apple a commission, 12% or 27%, even if it directed customers to the web.

Even though Epic filed the case and it wasn’t a class action, the injunction was applied to all developers based in the United States. Apple clearly planned to appeal that point even then, but then things were made more complicated.

Epic filed a complaint, which resulted in Apple being found in contempt. However, the original injunction didn’t mention anything about Apple’s commission, and the violation was argued in spirit.

Various appeals and arguments later, and Apple has been told it is owed a commission, even on external links. The problem is, Apple would have to come back to court and decide on the commission rate with Epic.

That’s how the case has arrived at the Supreme Court. And even though Apple tried to get the proceedings halted in the lower courts, twice, it must now face both at once.

Apple’s arguments

The foundations of Apple’s arguments appear to be sound. The courts do appear to be ignoring the precedent set by CASA.

Epic’s iPod-like ad

The Supreme Court ruled that lower courts were exceeding their jurisdiction by applying injunctions outside the scope of a case. However, the 9th Circuit has argued that there is an antitrust exception to CASA that would allow the decision in Apple’s case to stand.

Apple believes very strongly that this effectively bypasses the Supreme Court’s ruling and authority. That’s why it said it would render the CASA case a dead letter.

The other argument also has to do with how the 9th Circuit does business. Apple argues that in the other Circuit Courts, civil contempt is applied only if the letter of the law is violated, not the spirit.

Even if you don’t care about any of this legal back and forth, it is still incredible that the Epic Games lawsuit has reached this point. It started with a “1984” parody ad starring an apple wearing sunglasses and could finish with setting incredibly important precedent via the Supreme Court.

If Apple wins the “in spirit” portion of its arguments, Apple gets to carry on with its previous 12% and 27% commission rates for external linking. It would also mean proceedings in the lower courts would return to appeals stages.

If the universal nature of the injunction is thrown out, then only Epic will be affected by Apple’s move away from anti-steering practices. It would mean a total and abject failure of a case that cost Epic over a billion dollars already.

Apple has requested that its petition be considered during the Supreme Court’s June 25 conference. Perhaps Epic’s CEO should hold off on celebrations until after that date.