Singapore AI News & Daily Briefing

Singapore is expanding its RoboNexus programme to help local robotics firms scale up and go global. The initiative connects startups with funding, mentorship, and partnership opportunities. RoboNexus now aims to position Singapore as a hub for robotics innovation, with a focus on industrial automation, logistics, and service robots.

A key market signal for AI spending is flashing warning signs. The Silicon Data LLM Token Expenditure Index, which tracks what businesses pay to use AI, has dropped nearly 20% from its May peak. While token prices have fallen dramatically since 2023, experts are divided: bulls say cheaper AI is expanding the market, while bears warn that falling willingness to pay could undermine the massive capital investment being poured into AI infrastructure. The US$700 billion-plus AI spending boom may be losing some momentum.

Microsoft is setting up a 6,000-person team to help businesses actually use AI tools, not just buy them. The workers will include engineers and industry veterans who can advise companies on deploying AI efficiently while cutting costs. This marks a shift from traditional software sales—now vendors are getting hands-on with implementation work. The move comes as businesses struggle with ballooning AI expenses.

Google DeepMind has partnered with indie film studio A24 in what they call a "first-of-its-kind" research collaboration. The partnership aims to explore how AI can support creative storytelling and filmmaking. Details on specific projects or timelines haven't been disclosed yet, but the collaboration signals growing interest from AI labs in the entertainment industry.

AI data centers are creating new headaches for power grids. Unlike regular factories, these facilities can suddenly spike electricity use within milliseconds when training or inference kicks in, throwing off utility planning and backup systems. The IEA estimates data centers could soon consume 3-4% of global electricity. Northern Virginia's 'Data Center Alley' is already straining local substations and transmission lines. The real problem isn't just how much power AI needs—it's how erratically it demands it, forcing grid operators to rethink their entire approach.

TikTok has laid off around 20 employees in Singapore from its Trust & Safety team, as part of a global restructuring affecting multiple countries including Indonesia, Malaysia, and Ireland. The company says it is centralising operations into key hubs and evolving how it works to keep teams scalable while advancing platform safety through new technology. Affected staff are being given support consistent with local laws.

Temasek's portfolio surged 14.8% to $401 billion, boosted by AI investments in Anthropic and OpenAI alongside strong Singapore-based holdings. The state investor plans to roughly double its AI allocation from 6% to 10-15% of its portfolio by 2031, spreading bets across the entire AI chain from data centers and energy to semiconductors and AI applications. Singapore remains Temasek's biggest market exposure at 27%, followed by the US at 26%.

The New York Times is asking a court to sanction OpenAI, accusing the AI firm of lying for two years about its ability to search ChatGPT logs while allegedly hiding samples containing billions of conversations. The Times claims OpenAI faked technical limitations to block discovery in the ongoing copyright lawsuit over training data. If the court agrees, OpenAI could face serious penalties that may damage its defense.

Singapore is implementing AI in healthcare with a focus on keeping humans in charge. The approach emphasizes that AI assists doctors rather than replacing them, with clear guidelines ensuring medical professionals oversee AI recommendations. This balances innovation with patient safety in hospitals and clinics.

Singapore is pushing for global teamwork on AI — saying countries must work together to make sure AI is developed responsibly and benefits everyone, not just a few nations. The government is calling for an inclusive approach where different voices are heard in shaping AI rules. This echoes Singapore's broader Smart Nation push, where AI adoption is being woven into public services and workforce planning.

D-Matrix, a Silicon Valley AI chip startup backed by Microsoft, has started shipping its Corsair inference accelerator. The chip is designed to work alongside Nvidia GPUs to speed up AI tasks with lower power use. Early tests show 2x to 10x better performance in certain coding workflows. The company has raised around $500 million at a $2 billion valuation.

Southeast Asian voice AI startups are competing against global players like ElevenLabs by training models on unusual datasets—including screaming, dog barks, and explosions—to detect fraud and voice manipulation. Singapore-based Wiz.ai and Indonesia's Kata.ai are deploying voice AI in banks and financial services for telemarketing and customer service automation.

A roundup of AI investment trends covering Temasek's increased AI portfolio allocation, a VC firm's focused AI strategy, and OpenAI's latest voice model capabilities. References earlier coverage of Temasek's AI investment plans.

South Korean memory chipmaker SK Hynix has overtaken Samsung Electronics to become Korea's most valuable company for the first time since 2000. SK Hynix shares rose 5.7% on June 22 while Samsung gained just 0.4%. The shift highlights how demand for high-bandwidth memory (HBM) chips — essential for AI training and running AI models — has favored SK Hynix, which supplies chips to Nvidia and Google. The company bounced back from a $5 billion loss in 2023 to a record $15.3 billion profit in 2024.

OpenAI is now showing ads to free ChatGPT users in South Korea, expanding a monetization test already running in the US, UK, Canada, Australia, and New Zealand. Pay subscribers won't see ads. The company says ad data is aggregated and won't include conversation content.

Serval, a US startup building AI agents to automate IT tasks like ticket management and employee onboarding, has raised $47 million in Series A funding from investors including Redpoint Ventures and General Catalyst. The company claims its AI already cuts IT resolution times by 90% at some clients like Verkada. While promising, there are no public metrics on paying customers or revenue, raising questions about whether this is real product-market fit or just early buzz.

Southeast Asian venture capital firms are shifting investment focus to AI startups, with some redirecting capital to US AI hubs while others seek local opportunities. Manila's Kickstart Ventures has turned fully toward AI companies since 2022 due to LP demands, backing San Francisco insurtech Lydia.AI and inference platform Featherless.AI. Singapore-based Blue Whale Energy exemplifies AI-adjacent startups attracting regional VC interest.

Australian data center operator Nextdc has secured $1.6 billion in debt financing, bringing its total available senior debt facilities to A$8.7 billion. The funding closes mid-July and will support infrastructure expansion across Australia, a key market for cloud and AI services in the region.

A weekly funding roundup shows China's Robot Era secured $148 million. The article is paywalled, so specific details about investors or the company's plans are not visible. TechInAsia's weekly digest tracks startup funding across Asia, giving a broad overview of investment trends in the region.

OpenAI and Broadcom have unveiled "Jalapeño," OpenAI's first custom chip built specifically to run AI models faster and use less energy. The chip, made with Celestica, is being tested with GPT models and targets deployment by end of 2026. This marks OpenAI's push to own more of its computing stack instead of relying solely on Nvidia.

Micron is buying a Taiwan chip plant for $1.8 billion and raising its total US investment commitment to $250 billion by 2035. The move signals a major push to expand DRAM production capacity, which could help ease memory shortages driven by surging AI demand for chips used in data centers and AI servers.

Meta is exploring a cloud business to sell AI computing power and hosted models, potentially competing with AWS, Microsoft Azure, and Google Cloud. CEO Mark Zuckerberg called it 'definitely on the table' given outside demand and possible excess data center capacity. The company plans $145 billion in capital spending this year.

South Korea's exports surged 29% year-on-year to $67.5 billion in February, with semiconductor exports jumping 160.8% to $25.2 billion. Korean memory giants Samsung and SK hynix are riding an AI-driven supercycle as high-bandwidth memory chips for AI data centers see record demand. PC makers face pressure as component costs climb and shipments could drop 5-9% this year.

Google is now labeling AI-generated or AI-modified ads on its Search and YouTube platforms, likely to comply with new advertising transparency rules. This means you'll start seeing notices on ads that use AI tools to create images, text, or video. It's part of a broader push by regulators worldwide to make AI use in advertising clearer to consumers.

Google is now adding AI labels to advertisements shown on its Search results and YouTube videos, making it clearer when ads contain AI-generated content. This move comes as regulators worldwide push for more transparency around synthetic media in advertising. Users will start seeing labels indicating when AI tools were used to create or modify ad content.

Osmond Chia - Business reporter Thu, 9 July 2026 at 7:59 pm GMT-7 2 min read South Korean computer chip maker SK Hynix has raised $26.5bn (£19.8bn) in its New York share offering, marking the largest ever listing by a foreign firm in the US. The company, a key supplier to artificial intelligence (AI) chip giant Nvidia, said on Thursday that it had sold 177.9 million American depositary shares for $149 each. The shares are set to begin trading on Friday on the Nasdaq. In May, SK Hynix saw its [market value top $1tn in its home country](https://www.bbc.com/news/articles/cnvp9dq0p3go?xtor=AL-72-%5Bpartner%5D-%5Byahoo.north.america%5D-%5Blink%5D-%5Bnews%5D-%5Bbizdev%5D-%5Bisapi%5D), lifted by the boom in demand for AI chips. Its share price has more than tripled in South Korea this year, which

The UN World Food Programme is testing remote-controlled SHERP vehicles equipped with AI to deliver aid in dangerous, hard-to-reach places like conflict zones. The AI-enabled trucks can be steered by humans from a safe distance through floods, swamps, and risky terrain. The UN refugee agency also showed a new AI legal assistant to help process refugee claims faster. Another UN project uses AI to analyze mobile phone data and satellite images to track where people flee during disasters, helping responders reach the right locations faster.

Tencent-backed Enflame Technology, a Shanghai-based AI chip developer, has won regulatory approval for an $888 million IPO on China's STAR Market. The chipmaker's revenue is expected to more than triple in the first half of 2026, with Tencent accounting for 84% of its 2025 sales. The listing is part of a wave of fundraising by Chinese AI companies seeking capital for AI cloud chip development.

Samsung Electro-Mechanics is planning a major investment to expand its Sejong plant with AI server packaging lines for advanced chip substrates. The move targets the growing demand for flip-chip ball grid array substrates used in AI accelerators and server CPUs. The global package substrate market is projected to grow 10% annually through 2027.

The UN's AI for Good summit in Geneva brought together tech executives, humanitarian workers, and governments to debate AI governance. Critics argue that big tech companies have too much influence over humanitarian AI projects, while others worry about global inequality as AI access concentrates among a few nations. Pro-Palestine activists also disrupted the event, highlighting how geopolitics is increasingly intertwined with technology debates.

Hong Kong startup Decisions Lab is using AI to simulate fake consumer personas, letting companies test marketing campaigns before spending real money. They claim 92.5% accuracy in predicting what real customers would do. One test with Hong Kong's government showed their AI correctly predicted 40% marketing waste, matching the actual 41.9% result. The startup plans to raise $1 million in mid-2026.

Deutsche Telekom, serving 300 million customers across Europe and the US, has become one of the world's first AI-native telecoms. The company rolled out ChatGPT Enterprise to over 50,000 monthly active users, seeing a 546% spike in AI tool usage since early 2026. Beyond customer service chatbots, they're embedding AI into network operations—optimizing tower performance in real time as crowds shift throughout the day. Their most ambitious move: bringing AI directly into phone calls, with features like live translation and post-call summaries built into the voice network itself. The lesson for other companies? Treat AI as a total redesign of how work gets done, not just another software install.

Singapore's Temasek is planning to nearly double its direct AI investments, targeting 10-15% of its S$518 billion portfolio by 2031—worth roughly S$52-78 billion at current values. The state investment firm says the other 85% of its portfolio must also adopt AI to stay competitive. Its AI holdings span NVIDIA, TSMC, Anthropic, OpenAI, and cloud providers, plus Temasek-created firms like Temus and Aicadium. CEO Dilhan Pillay stressed that AI replaces tasks, not jobs, and that reskilling workers is crucial.

Nearly half of Singapore employees—49%—expect their jobs to be affected by automation, restructuring, or cost-cutting in 2026, a figure far higher than the global average of 37%. The finding comes as larger employers become more cautious about hiring, with retrenchments climbing to their highest quarterly level since 2023. Most workers responding to the insecurity are already taking action: 85% say they'd start job-hunting, while 64% plan to invest in new skills. Singapore also stands out for being one of the world's most office-centric workforces, with 37% of employees working five days in-office—double the global average.

Chinese AI startup MiniMax is seeking up to $2 billion in funding to fuel its expansion. The company's founder is going to extreme lengths to show confidence—saying he'll skip his salary until the firm achieves artificial general intelligence. It reflects how hot AI startups are chasing big money right now.

Tesla is rolling out its Full Self-Driving feature in South Korea, starting with US-made Model 3 and Model Y vehicles. The update is coming in stages. For South Korean drivers, this means access to Tesla's advanced driver-assist features, though it still requires human supervision.

The IMF says South Korea, Taiwan, Thailand, and Malaysia are outperforming the rest of the world thanks to AI. These four economies are the biggest net exporters of AI hardware—think memory chips and processors—and they've rode the global AI boom to become standout economic performers. The finding comes from the IMF's July 2026 World Economic Outlook, which highlights how countries with deep semiconductor supply chains are pulling ahead economically.

Malaysian Prime Minister Anwar Ibrahim is launching an AI version of himself to interact with the public. Built by local firm Zetrix AI, the avatar called PMX AI has been trained on Anwar's speeches, writings, and government policies to mimic his voice and likeness. What's notable is it's an 'agentic' AI—it can break down tasks and execute them autonomously without constant human oversight. The PMO has already released a launch video introducing it. It's among the first times a head of government has deployed a digital AI twin for public engagement.

OpenAI launched GPT-5.6 for coding, cybersecurity, science, and work. GPT-5.6 adds new reasoning modes, multi-agent workflows, and stronger safety testing. OpenAI has made its GPT-5.6 model family generally available following a limited preview, launching three models for different performance and cost requirements: Sol, Terra, and Luna. Sol is the flagship model, Terra is intended for […] The post OpenAI launches GPT-5.6 for coding, cyber and science appeared first on TechWire Asia .

A new app called Ode uses AI to generate personalized poetry as a mood-lifting tool, with British actor Stephen Fry's voice reading the verses. Think of it as a digital wellness companion that serves you poems on demand based on how you're feeling. It's part of a growing trend of AI-powered mental health tools, though this one leans on literature rather than traditional therapy.

A new HSBC survey finds 76% of Singapore investors now use AI tools for finance and investing — above the global average of 72%. High-net-worth Singaporeans are leading the pack at 90% adoption. But here's the twist: most still bring AI insights to human advisers before making final calls. Only 8% say AI alone was the biggest influence on their last major investment. The takeaway? Singaporeans see AI as a research assistant, not a replacement for professional advice.

Singapore-headquartered Bitdeer — a Nasdaq-listed firm providing AI computing and Bitcoin mining infrastructure — has broken ground on a $36 million electronics factory in Nevada. The 187,000-square-foot plant will churn out 10,000 computing hardware units monthly and create 70 local jobs. It's Bitdeer's first US manufacturing base, designed to shorten supply chains and serve American customers faster.

Cybersecurity firm Kaspersky uncovered a multi-stage phishing campaign since April 2026 targeting factories in Malaysia, Egypt, Czech Republic, and Russia. Attackers pose as buyers requesting pricing, then redirect targets to fake login pages stealing corporate credentials. Kaspersky notes attackers increasingly use AI to craft convincing phishing emails, making schemes harder to spot.

OpenAI has released GPT-5.6, a family of three AI models (Sol, Terra, Luna) designed for coding, cybersecurity, and scientific work. The models include new reasoning capabilities, the ability to work together on complex tasks, and stronger safety testing before deployment.

This opinion piece argues OpenAI's vision of 'freeing humans from screens' conflicts with its business model of keeping users engaged with AI products. It questions whether OpenAI can truly liberate users when engagement drives revenue.

A new study using an AI-detection browser extension found that up to 41% of long posts on LinkedIn and roughly a third of articles on X are fully AI-generated. The research suggests humans are regularly wading through AI-generated slop on major platforms without realizing it. LinkedIn says it's trying to reduce AI content, while Reddit has launched ads emphasizing its human users.

Singapore's affluent investors are leading global AI adoption in finance—76% use AI tools for investment research, rising to 90% among those with over $2 million in assets. Surprisingly, older generations (Gen X and Baby Boomers) are driving this trend locally, outpacing global peers. However, most still prefer a hybrid approach: AI for research and stress-testing, then human advisers for final validation. HSBC Singapore has launched its own AI adviser tools and partnered with Google Cloud to roll out over 200 AI use cases.

Singapore workers are the least skeptical of AI globally (only 29% are skeptics vs 37% worldwide), yet they use AI least in their daily work — just 6% use it regularly, half the global average. The reason? Corporate AI rollouts keep failing. Forty percent cite generic, unhelpful outputs as the top reason, the highest globally. The key to fixing this is grounding AI in company-specific data and integrating it into existing workflows. Singapore insurer Singlife is cited as a success story, using Salesforce's Agentforce platform grounded in their own product manuals.

Singapore workers are the least skeptical about AI globally — only 29% express doubts, compared to 53% in the US and UK. But here's the irony: just 6% actually use AI daily in their work, nearly half the global average of 11%. The gap isn't about resistance — it's about disappointment. When Singapore companies tried AI pilots, 40% failed because outputs were too generic, the highest failure rate globally. Insurer Singlife cracked the code by training AI on its own product manuals, letting customer service staff get instant, accurate answers instead of manual document searches.

Large companies in Southeast Asia are struggling to defend against AI-powered cyberattacks with understaffed security teams. Kaspersky detected over 18 million malicious attacks across SEA in 2025, with Singapore alone facing over a million attacks. The problem: AI lets hackers launch attacks faster and more convincingly, while many organisations still use dozens of disconnected security tools that create blind spots and overload analysts with alerts. Experts say companies need to consolidate their security systems and use AI-driven automation to keep up, shifting from reactive firefighting to proactive threat hunting.

The UK government is now treating Microsoft, Google, Amazon, and Oracle as "critical third parties" to the financial system, putting them under direct regulatory oversight. The concern: as banks and insurers increasingly rely on cloud services, an outage at a major provider could simultaneously knock out multiple financial firms. The designation takes effect July 13.

The filing reflects a historical equity investment completed in 2025, rather than a new funding round.

Refik Anadol opened Dataland, billed as the world's first AI art museum, in Los Angeles. Visitors wear biosensors that let the gallery's AI respond to their biometrics, creating personalized, ever-changing digital environments. Anadol argues this proves AI can be a legitimate creative tool rather than just 'AI slop'. The museum deletes all visitor data once they leave.

OpenAI has launched ChatGPT Work, its long-awaited 'super app' that bundles the chatbot, coding tools, and new models to handle work tasks both independently and collaboratively with users. It launched alongside new GPT-5.6 model tiers, as OpenAI separately works on building a fully automated researcher.

The European Union has found Meta in preliminary breach of its Digital Services Act for designing Facebook and Instagram with "addictive" features—specifically infinite scroll, autoplay videos, and push notifications that keep users hooked, including minors. Meta could face fines up to 6% of its global annual turnover. The EU wants features disabled by default and screen time breaks enforced.

Tencent is in talks to become the largest shareholder of Manus AI, stepping in after Beijing ordered Meta to unwind its $2 billion acquisition of the Chinese AI startup. Tencent will reportedly buy Manus for no less than $2 billion, giving it significant influence over one of China's rising AI players.

[](https://bit.ly/3qgqzHg) [](https://bit.ly/3KjTj94) A paddleboarder died following a lightning strike at the Pasir Ris beach area on Jul. 5 at about 4:46pm. In response to queries from _Mothership_, the Meteorological Service Singapore (MSS) said the lightning strike was detected when the main thunderstorm cloud was still a few kilometres away. ## ADVERTISEMENT A total of 1,569 lightning strikes were observed that day, MSS added. ## Thunderstorms developed near Yishun According to MSS, thunderstorms that day began developing in the north of Singapore near the Yishun area at about 3:20pm and moved north-westwards. Forecasts for thundery

Meta's new AI image tool built into Instagram can now use your public photos to generate new images without your permission. All someone has to do is tag your profile. SAG-AFTRA and talent agency CAA are urging users to opt out through Instagram's settings under 'Sharing and reuse.' You won't be notified if your likeness is used.

Bun, the popular JavaScript runtime and toolkit, rewrote its entire codebase from Zig to Rust in just 11 days — with Anthropic's Claude Fable 5 AI doing most of the heavy lifting, producing over a million lines of new code. The move aims to improve Bun's performance and developer experience. The project highlights how far AI has come in handling large-scale, real-world coding tasks.

Digital AI like ChatGPT works in two dimensions — on screens. But the next frontier is "embodied AI" — robots that move and act in the physical world. This requires training on real-world physics data, not just text, and that's much harder. Robots that score 90% in simulations often drop to 30-60% in reality. Companies racing to collect physical data first will have an advantage. Nvidia's Jensen Huang says young people today should focus on physical sciences — not just coding — for tomorrow's opportunities.

OpenAI and Google reportedly sold AI models to blacklisted China-based groups through Singapore subsidiaries of Alibaba, Baidu, and Tencent. This raises serious questions about how US technology restrictions are being enforced—and bypassed—as AI exports face increasing scrutiny amid US-China tech tensions.

The era of 'learn to code and get a tech job' is over. With AI writing code now, entry-level programming jobs have dried up — and reskilling programs like Code Louisville are shutting down. Companies like Accenture are cutting workers who can't adapt to AI tools. But the burden is shifting to employers: 77% of firms globally plan to upskill workers, and groups backed by OpenAI and Anthropic are pushing for wage insurance for displaced workers. Entry-level roles are becoming the most endangered.

Anthropic researchers developed a tool called 'J-lens' that uncovers a hidden area inside Claude they call the J-space. This space contains words related to the model's response that it may not ultimately produce—in essence, what Claude is 'thinking' before it speaks. It's a breakthrough in understanding AI's inner reasoning processes.

The US Federal Reserve has appointed venture capitalist Marc Andreessen to advise on AI's economic impact. Fed Chair Kevin Warsh sees AI as a potential "disinflationary force" that could help tame inflation. However, Andreessen's firm Andreessen Horowitz has major investments in AI companies, raising red flags about conflicts of interest in his advisory role.

OpenAI has shut down its Atlas browser just eight months after launch, moving its features into the ChatGPT Chrome extension. It's the latest in a string of OpenAI products that haven't taken off. The company is apparently pivoting to integrate AI directly into existing tools people already use rather than creating standalone products.

Meta has launched a new AI tool called Muse Image on Instagram that lets anyone create AI-generated images using photos from public accounts. Hollywood's major talent union SAG-AFTRA and agency CAA are pushing celebrities to opt out, warning it enables unauthorized use of their likenesses. Cybersecurity firms Malwarebytes and Proton flagged the feature as a privacy risk, noting the opt-out option is buried deep in settings and could enable scams and impersonation. Users under 18 are automatically excluded.

SpaceX has asked US regulators for permission to launch 100,000 next-generation Starlink satellites — nearly 10 times the current constellation. The company promises these Gen3 satellites will deliver gigabit internet speeds with latency under 20 milliseconds, aiming to support "billions of AI-powered devices worldwide." However, the satellites would weigh over 2 tons each, requiring SpaceX's not-yet-ready Starship rocket. Astronomers are already protesting, warning the massive constellation would "devastate" ground-based astronomy.

The European Union has ordered Meta to overhaul Facebook and Instagram's addictive features, including infinite scroll, autoplay videos, and engagement-focused algorithms, or face fines up to 6% of global revenue. The EU says these features are designed to trap users in "autopilot mode," harming mental health especially among teenagers. Meta must now disable key features and prove its platforms are safer for users.

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

SK Hynix, the world's top maker of AI memory chips (HBM), saw its shares surge 14% in its Nasdaq debut after a $26.5 billion share sale. The IPO — the second-largest in U.S. history — shows investors remain hungry for semiconductor stocks tied to the AI boom, despite recent market volatility. SK Hynix's chips are essential components in Nvidia and AMD AI processors.

A political commentary draws parallels between climate activism strategies and AI development, arguing that the AI industry should focus on building independent power structures rather than lobbying existing power players. The piece reads more like an activist essay than substantive AI analysis, offering strategic advice for how AI development could mirror grassroots movements in gaining influence. The actual content appears thin, with the article URL suggesting it's more political manifesto than news.

Google DeepMind researchers have identified six ways hackers can hijack AI agents - programs that browse the web and complete tasks autonomously. These "AI Agent Traps" include hidden instructions invisible to humans, manipulation through persuasive language, distorted memory, overridden safety features, coordinated multi-agent attacks, and tricks that fool human reviewers approving AI actions. The researchers warn that the open internet can be weaponized against AI systems and recommend stronger defenses, runtime scanners, and clearer legal accountability.

Tencent is in talks to buy a majority stake in AI agent startup Manus at the same $2 billion valuation that Meta was forced to abandon after Beijing blocked the deal. The move aligns with Tencent's own AI agent ambitions, including integration with WeChat.

Meta now lets Instagram users in most countries stop their photos from being used to train AI tools. The setting is buried in privacy options — go to Settings > About > Privacy Policy, then find the AI training toggle. It's opt-out, not opt-in, which means your content could train AI unless you actively say no. This matters because AI companies have been scraping social media data for years, often without clear user consent.

Something went wrong [Skip to navigation](https://sg.finance.yahoo.com/news/apple-files-lawsuit-accusing-chatgpt-205349400.html#navigation-container)[Skip to right column](https://sg.finance.yahoo.com/news/apple-files-lawsuit-accusing-chatgpt-205349400.html#right-rail) # Yahoo Finance Tip: Try a valid symbol or a specific company name for relevant results ### Trending Tickers * S SKHY 0.00 (0.00%) * D D05.SI 70.45 (+0.31%) * U U11.SI 44.38 (+0.09%) * Z Z74.SI 4.4000 (+0.2278%) * S SE 111.14 (+1.86%) * [News](https://sg.yahoo.com/) * [Finance](https://sg.finance.yahoo.com/) * More * [News](https://sg.yahoo.com/) * [Today's news](https://sg.news.yahoo.com/) * [Singapore](https://sg.news.yahoo.com/singapore/) * [Politics](https://sg.news.yahoo.com/politics/) * [Southeast Asia](https://sg.news

Don't want someone using your Instagram pics to generate AI content? Here's what to do. A 1x1 image, likely be a tacker probe

Apple has sued OpenAI, claiming the AI company stole trade secrets by recruiting Apple engineers to bring confidential iPhone and hardware designs to job interviews. Apple says OpenAI's new hardware chief, who spent 24 years at Apple, coached departing employees on how to bypass Apple's security systems and even asked candidates to bring actual iPhone parts to interviews. OpenAI is reportedly building a smartphone that could compete directly with the iPhone.

Security researchers have documented the first known AI-powered ransomware attack, called JadePuffer, which used an AI agent to automate much of a cyber extortion operation. The attack used an AI assistant to handle reconnaissance, credential theft, database encryption, and ransom note creation, dramatically reducing the human labor needed. Experts warn this marks a turning point where AI makes ransomware cheaper, faster, and easier to execute at scale, potentially leading to a surge in automated attacks. Companies are urged to secure their AI development tools from public internet exposure.