News World That's Worth

NVIDIA Reduces Financial Guarantee for OpenAI’s Ohio Data Center Project to Under $12 Billion According to a report from The Wall Street Journal , NVIDIA has significantly scaled back its financial guarantees for OpenAI’s massive data center expansion in Ohio. The commitment has been reduced from an initial $250 billion to under $12 billion following growing pushback from investors concerned about the chipmaker's exposure to massive balance sheet liabilities. The adjustment comes as OpenAI remains in active negotiations to secure a legally binding lease for the 10-gigawatt (GW) facility. Sources familiar with the matter indicate that NVIDIA and OpenAI are nearing a finalized agreement that limits NVIDIA's direct financial backing exclusively to the project's first phase. The formal agreement could be signed as early as this weekend, with final terms expected to be locked in by August 17 . The site is being developed by SB Energy , a subsidiary of SoftBank Corp. Once full...

Apple Reportedly Training Custom In-House AI Model for China in Collaboration with Alibaba According to a report from Reuters citing sources familiar with the matter, Apple has begun training a custom, localized artificial intelligence model specifically designed for the Chinese market. The initiative is being developed in partnership with Alibaba , which is providing technical and infrastructure support. This development marks a strategic shift from earlier reports, which suggested Apple would directly integrate pre-existing models from Chinese tech giants Alibaba and Baidu to power Apple Intelligence on devices sold in mainland China. Currently, Apple Intelligence remains unavailable in China one of Apple's most critical global markets. However, regulatory hurdles appear to be clearing; reports last month indicated that Chinese regulatory authorities approved an Apple Intelligence framework backed by Alibaba’s Qwen foundation model. An official regional launch may be imminent f...

Anthropic Reaches Q2 2026 Profitability Milestone with Revenue Surging 14x to $11.5B According to a financial update delivered to investors, Anthropic posted extraordinary results for the second quarter of 2026, generating $11.5 billion in revenue . This represents a 14-fold (1,360%) increase compared to the $787 million recorded in the same period last year meaning Anthropic generated more revenue in Q2 alone than in all of 2025 combined. Crucially, Anthropic reported its first-ever positive adjusted net operating profit, reaching $559 million for the quarter. Looking further ahead, the investor documents outline a long-term projection forecasting annual revenues to reach $190 billion to $200 billion by 2028 , underscoring the massive growth runway management anticipates for enterprise AI adoption. These forward-looking metrics are heavily influencing Anthropic’s valuation trajectory. As the company weighs future funding rounds or an eventual initial public offering (IPO), financia...

Alibaba Cloud Launches Qwen3.8-27B: Local GPU-Ready AI Model Matching Claude Opus 4.6 in Coding Alibaba Cloud has officially released Qwen3.8-27B , its latest compact artificial intelligence model optimized to run locally on consumer-grade graphics cards while delivering a substantial leap in performance over its predecessor, Qwen3.6-27B. Despite its modest parameter size, Qwen3.8-27B achieves coding capabilities comparable to Anthropic’s flagship Claude Opus 4.6 . Benchmark results highlight significant performance gains: Terminal Bench 2.1: Reached a score of 73.0 . SWE-bench Pro: Advanced to 61.7 . DeepSWE: Rose to 42.2 . Alibaba has made the model weights available for public download and confirmed that an official first-party API service will launch at a later date. However, third-party hosting is already live on OpenRouter via AkashML , priced at $0.45 per million input tokens and $3.20 per million output tokens  a rate considered relatively high given the model's footp...

Microsoft Unifies Consumer and Commercial Copilot Apps into a Single Platform Microsoft has officially begun consolidating its two parallel AI applications   Copilot (aimed at consumers) and Microsoft 365 Copilot (tailored for enterprise users) into a unified experience across desktop, mobile, and web platforms. Under the new strategy, the enterprise-grade Microsoft 365 Copilot app will serve as the core foundational codebase. Moving forward, it will simply be rebranded as Copilot , accompanied by a refreshed icon design signaling the update. The unified application will support dual authentication, allowing users to sign in with both personal Microsoft accounts and enterprise credentials simultaneously while keeping all data and chat histories strictly segregated. As part of the consolidation effort, Microsoft is streamlining the feature set. Effective August 18, 2026 , legacy features including Group Chat , Podcasts , and Deep Research will be officially retired from the uni...

Gemini Introduces Toggle to Remove Visible Watermarks from AI-Generated Media Google Gemini has officially announced a new setting that allows users to disable the visible watermark  the signature starburst icon located in the bottom-right corner on AI-generated images, videos, and music tracks. The new configuration toggle is located in the Gemini settings menu under a dedicated category titled "Media watermark," which offers a simple On/Off switch. In addition to the main Gemini application, Google's AI video generation app, Flow , will also feature this opt-out setting. Crucially, this preference only affects human-visible visual indicators. Invisible provenance tracking measures—including Google's SynthID digital watermarking and standardized C2PA metadata tags will remain embedded in all generated media files to preserve content authenticity.

Anthropic Reveals Internal Frontier AI Models 'Model 1' and 'Model 2' in August 2026 Risk Report In its August 2026 Risk Report, Anthropic disclosed the existence of two unreleased internal artificial intelligence architectures: Model 1 and Model 2 . According to the report, Model 1 closely mirrors the performance profile of Claude Mythos 5 with minor structural differences. However, Model 2 demonstrated a significant capability leap, described in the report as being "somewhat more capable" than Mythos 5. Anthropic revealed that Model 2 is already being used heavily alongside Claude Mythos 5 across internal workflows, including software engineering, advanced research, and cybersecurity vulnerability assessment. Escalating Safety Risk Tiers and Evaluation Limitations The advanced capabilities of both Claude Mythos 5 and Model 2 prompted Anthropic's safety teams to re-evaluate potential threat vectors particularly risks associated with autonomous system ...

Google Upgrades Android Quick Share with Proximity-Based Contact and Media Transfer Google has officially released a major update to Android Quick Share , introducing seamless proximity-based sharing capabilities. Users can now exchange contact information, photos, and videos simply by bringing two compatible Android smartphones close to one another. When two devices are placed side-by-side, an automated prompt appears on-screen asking users to choose which contact details such as profile photos, phone numbers, or email addresses they wish to share for an instant mutual exchange. For media transfers, the physical proximity trigger bypasses manual device scanning, allowing users to send photos and videos directly through the standard Quick Share pipeline. Supported Hardware & Rollout Schedule The proximity sharing feature is currently live for Google Pixel 6 models and newer. Google confirmed that support will next expand to Samsung’s latest foldable flagship lineup the Galaxy Z Fo...

OpenAI Introduces 'Computer History' for ChatGPT on macOS to Track and Recall User Workflows OpenAI has rolled out a major update for the ChatGPT macOS app, introducing a new feature called Computer History . Designed to replace the earlier experimental "Chronicle" extension, this feature allows ChatGPT to log on-device activities such as clicks, keystrokes, and application usage. The goal is to provide the AI with contextual awareness of user workflows, enabling it to answer historical queries like, "What was I trying to debug yesterday?" Addressing potential privacy concerns upfront, OpenAI has designed Computer History to be opt-in by default . Users must manually enable the feature and specifically whitelist which applications and websites the AI is permitted to track. OpenAI also explicitly clarified that the system only logs telemetry and activity data; it does not record screens or capture audio . The feature is integrated into both the standard ChatG...

Z.ai Launches Open-Source GLM-5.3: Enhanced Coding, Cybersecurity Upgrades, and Dynamic API Pricing Z.ai has officially released GLM-5.3 , the latest iteration of its open-source artificial intelligence model. Built upon the same foundational architecture as its predecessor, GLM-5.2, this new version leverages advanced post-training techniques to significantly elevate its performance across multiple domains. A major focus of GLM-5.3 is its enhanced programming capabilities. According to Z.ai, the model scored 50% higher than GLM-5.2 on internal coding benchmarks. It also demonstrated outstanding performance on external evaluations such as Terminal Bench 3.0 and Agents' Last Exam, positioning it as a highly competitive alternative to leading proprietary models from OpenAI and Anthropic. Reflecting current industry demands, GLM-5.3 also introduces robust cybersecurity features. In CyberGym testing environments, the updated model successfully identified more than twice as many secur...

Google has officially rolled out Gemini 3.7 Flash , featuring massive performance upgrades particularly in software engineering and coding workflows that bring it dangerously close to top-tier models like Anthropic's Claude Sonnet 5 and OpenAI's GPT-5.6 Terra . To celebrate the launch, Google has announced a promotional 50% price cut running through the end of the year, making Gemini 3.7 Flash one of the most cost-effective and high-value frontier models on the market. On the Artificial Analysis Intelligence Index , Gemini 3.7 Flash scores a solid 56 , sitting just one point behind Meta’s Muse Spark 1.2 while outperforming the latest DeepSeek V4 Pro (0813) . Benchmark tests show that Gemini 3.7 Flash competes tightly in production code quality and long-horizon software engineering evaluations, proving that Google's Flash tier can now go toe-to-toe with heavier mid-tier competitor models at a fraction of the cost. Arriving just weeks after Gemini 3.6, this release highlig...