Homechevron_rightArchiveschevron_rightTechnology

Archives

Explore Stories by Date

14 Jan 2026

Phison aiDAPTIV+ delivers 10X faster AI inference on consumer PCs using NAND flash as memory

Phison showcased aiDAPTIV+ at CES 2026, a software and hardware solution that accelerates AI inference by up to 10 times on consumer PCs. The technology uses NAND flash as memory expansion, enabling systems with just 32GB of DRAM to run AI models requiring 96GB through conventional approaches. Partners including Nvidia, AMD, MSI, and Acer demonstrated the technology, which addresses the memory bottleneck limiting AI deployment on ordinary hardware.

3 Sources

Nvidia releases DLSS 4.5 with enhanced AI upscaling as 6x frame generation waits until spring

Nvidia launched DLSS 4.5, introducing a second-generation transformer model for AI upscaling that delivers sharper textures and reduced ghosting in games. While the new Model L and M presets show measurable improvements in performance modes, the highly anticipated 6x Multi Frame Generation capability won't arrive until spring 2026. Testing reveals DLSS 4.5 makes the biggest difference for mid-range GPUs running at 1440p resolution.

4 Sources

Critical AI/ML Vulnerabilities Discovered in Python Libraries From Nvidia, Salesforce, and Apple

Security researchers uncovered vulnerabilities in popular Python libraries used across AI and ML models with millions of downloads. The flaws allow attackers to hide malicious code in model metadata that executes automatically when loaded. Nvidia NeMo, Salesforce Uni2TS, and Apple FlexTok were all affected, with patches now released following responsible disclosure.

2 Sources

Alibaba Links Taobao Shopping to Qwen AI App, Advancing Push Into Consumer AI Services

Alibaba is connecting its flagship Taobao shopping platform, Alipay, and travel services to its Qwen AI app, aiming to create a unified AI personal assistant for over 100 million users. The integration marks the company's most ambitious step yet in consumer-facing AI, as it seeks to demonstrate AI's money-making potential while leveraging its 35.8% share of China's AI cloud market.

4 Sources

Roblox's AI-powered age verification system stumbles with major accuracy and safety failures

Roblox rolled out AI-powered age verification globally to protect young users from predators, but the system is misclassifying ages and can be easily fooled. Users report being placed in wrong age groups, while age-verified accounts are being sold on eBay for as little as $4, raising serious questions about the platform's child safety measures.

5 Sources

Google Translate tests AI-powered feature to offer multiple translation choices

Google Translate is testing a new Show alternatives button that uses Gemini AI to provide up to three translation variants with different formality levels and phrasing. Discovered in app version 10.0.36, the feature addresses the limitations of literal translations by offering contextual options, particularly useful for languages with honorifics and idioms.

2 Sources

Docusign AI Contracts Tool Translates Legal Jargon Into Plain English for Faster Signing

Docusign launched AI-powered features to help users understand complex contracts before signing. The tool translates legal jargon into plain English summaries and answers specific questions about agreements. It also automates document preparation for creators, aiming to reduce signer hesitation and boost completion rates.

3 Sources

Teachers Union Exits X After Grok AI Generates Sexualized Images of Children

The American Federation of Teachers is leaving X after the platform's Grok AI chatbot produced sexualized images of minors without adequate safeguards. AFT President Randi Weingarten called the AI-generated content the "last straw," citing ongoing concerns about extremist activity and child safety on the platform since Elon Musk's 2022 takeover.

2 Sources

13 Jan 2026

DeepSeek's Engram tackles AI compute waste with conditional memory breakthrough

DeepSeek unveiled Engram, a conditional memory technique that separates static information retrieval from complex reasoning tasks. The method allows AI models to bypass GPU memory constraints by committing knowledge to system RAM, reducing reliance on expensive high-bandwidth memory while improving performance on long-context queries.

5 Sources

Signal founder Moxie Marlinspike launches Confer, an encrypted AI chatbot alternative to ChatGPT

Moxie Marlinspike, creator of Signal, has launched Confer—an open source AI chatbot that encrypts user conversations so thoroughly that even server administrators can't access them. Unlike ChatGPT or other mainstream AI assistants, Confer uses passkeys, Trusted Execution Environments, and remote attestation to ensure prompts and responses remain private, addressing growing concerns about data collection in AI platforms.

7 Sources

Google Veo 3.1 now generates vertical videos from reference images with 4K upscaling

Google updated its Veo 3.1 AI video generation model to create native vertical videos using reference images, targeting social platforms like YouTube Shorts and TikTok. The update enhances character consistency across clips and adds 4K upscaling capabilities. Users can now upload up to three images to guide video creation in the 9:16 aspect ratio, eliminating the need for cropping.

14 Sources

Apple Creator Studio locks new AI features behind subscription, leaving app owners in the lurch

Apple launched its Creator Studio subscription service bundling professional creative apps at $12.99 per month. But the move introduces a controversial shift: existing owners of apps like Final Cut Pro and Pixelmator Pro will no longer receive all new features unless they subscribe. The company plans to reserve AI-powered tools and premium content exclusively for subscribers, effectively creating a two-tier system that has sparked concern among longtime users who paid hundreds for standalone software.

5 Sources

Apple to begin mass production of in-house AI server chips in H2 2026 as Google deal buys time

Apple plans to start mass production of custom AI server chips in the second half of 2026, with proprietary data centers expected in 2027. Analyst Ming-Chi Kuo reveals the move aims to reduce reliance on partners like Google, whose Gemini deal is temporary. The chips, codenamed Baltra, mark Apple's push for control over core AI technologies.

4 Sources

WitnessAI raises $58 million to solve the multi-billion AI security problem enterprises face

WitnessAI just raised $58 million to build what it calls a 'confidence layer for enterprise AI' as companies deploy AI-powered chatbots, agents, and copilots across operations. The startup addresses how enterprises can use powerful AI tools without accidentally leaking sensitive data, violating compliance rules, or opening the door to prompt-based injections—risks that could turn AI security into an $800 billion to $1.2 trillion market by 2031.

3 Sources

Tamil Nadu partners with Sarvam AI to build ₹10,000 crore Sovereign AI Park in Chennai

Tamil Nadu signed an MoU with Sarvam AI to establish India's first full-stack Sovereign AI Park with an initial investment of ₹10,000 crore. The project aims to create a state-controlled AI ecosystem in Chennai where data, models, and computing infrastructure remain within trusted boundaries, generating around 1,000 high-skilled jobs in AI research and infrastructure engineering.

3 Sources

Get all updates into your inbox 💌, Stay Up-to-date for Free 🥳

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

The Outpost

News

About