AI news digest — Friday, August 14, 2026

Vulnerability giving attackers full control of Macs is under active exploitation

Industry · United States · Ars Technica AI

A newly disclosed flaw in macOS’s built‑in Screen Sharing service allows attackers to gain full remote control of a Mac without needing a password. The vulnerability, which affects recent versions of the operating system, is already being exploited in the wild, enabling threat actors to log in, steal data, install malware, or pivot to other systems. Apple has issued a security update, but machines that remain unpatched or have Screen Sharing left enabled are at immediate risk.

What this means for you: Users should immediately install the latest macOS security patch, disable Screen Sharing when not needed, and enforce strong authentication or network‑level restrictions to mitigate the risk.

PBS station fears losing 50TB of data after being ghosted by cloud storage provider

Industry · United States · Ars Technica AI

A PBS affiliate warned that it could lose approximately 50 terabytes of archival footage after its cloud storage provider, Iron Mountain, reportedly cut off access to the servers holding the data. The station says it has been unable to retrieve its files despite repeated attempts, raising concerns about permanent loss of valuable public‑media content. The incident highlights the dangers of relying on a single third‑party vendor for long‑term preservation without verified backup or data‑portability guarantees.

What this means for you: Organizations should maintain multiple, geographically dispersed backups, regularly test restore procedures, and negotiate clear SLAs and data‑exit clauses with any third‑party storage provider.

Mark Zuckerberg has an Instagzam

Products · United States · The Verge AI

Instagram unveiled a revised wordmark this week, replacing its long‑standing logo with a design that many users find difficult to read and no longer clearly spells the platform’s name. The change appears to be part of a broader effort to modernize the brand’s visual identity, though the rationale behind the specific tweaks has not been explained publicly. Design experts note that such alterations can affect brand recognition and user trust, especially when the new mark deviates significantly from the familiar iconography that has accompanied the app for years.

What this means for you: Marketers and designers should audit any external materials that use the Instagram logo to ensure they reflect the updated brand assets and avoid confusion.

You can now turn off Google Gemini’s visible watermarks

Products · United States · The Verge AI

Google has added a toggle in its Gemini AI suite that lets users disable the visible "sparkle" watermark applied to images, videos, and music generated by the service. When the setting is turned off, the conspicuous mark in the bottom‑right corner disappears, although Google says an invisible metadata tag remains to help identify AI‑created content. The move gives creators more control over the appearance of their outputs but raises questions about attribution and the potential for undisclosed AI use.

What this means for you: Before hiding the watermark, creators should consider disclosure requirements and platform policies to maintain transparency about AI involvement.

Google will now allow users to remove visible watermark from its AI generations

Industry · United States · TechCrunch AI

Google announced that users can now turn off the visible watermark on content produced by its AI tools, including Gemini and the Flow video generator. The update provides a simple switch to remove the on‑screen sparkle badge, while confirming that an imperceptible identifier persists within the file to support detection of AI origin. The change aims to give artists and developers flexibility in how their AI‑generated work is presented, though it also underscores the ongoing debate over visible versus invisible attribution methods.

What this means for you: Professionals should verify that removing the visible mark does not violate any licensing or attribution obligations tied to the AI service they are using.

Does Mark Zuckerberg really believe AI is ‘for everyone’?

Industry · United States · TechCrunch AI

Meta released Glimmer this week, an open-weight AI model anyone can download and run on their own hardware — a contrast to Muse Spark, the company’s more powerful model that stays locked behind its own APIs. The release landed alongside a letter from Mark Zuckerberg arguing AI should be “for everyone” rather than controlled by a handful of labs, but as Equity’s […]

Kog is going deeper to squeeze more inference out of GPUs

Industry · United States · TechCrunch AI

The idea that GPUs are poorly suited for agentic workflows may be a misconception, according to French startup Kog.

OpenAI and Anthropic in price war as Chinese AI rivals gain ground

Industry · United States · Ars Technica AI

US groups release cheaper models after new challenges to their trillion-dollar ambitions.

Hyperscalers might regret embracing natural gas if new forecast proves correct

Industry · United States · TechCrunch AI

Natural gas prices could triple in some parts of the U.S., which could saddle hyperscalers with massive bills to power their AI data centers.

Meta’s ‘open’ AI, and a $250M deal gone very wrong

Industry · United States · TechCrunch AI

Meta released Glimmer this week, an open-weight AI model anyone can download and run on their own hardware — a contrast to Muse Spark, the company’s more powerful model that stays locked behind its own APIs. The release landed alongside a letter from Mark Zuckerberg arguing AI should be “for everyone” rather than controlled by a handful of labs, but as Equity’s […]

Samsung health AI models analyse wearable biosignal data

Industry · United Kingdom · AI News

Samsung Research America’s Digital Health Team has presented two AI foundation models designed to learn from wearable biosignals. The work centres on data captured by smartwatches, including heart activity, sleep, and physical activity. The company discussed its Connected Care vision at the Health Forum during Galaxy Unpacked in July 2026. Samsung described a future of […] The post Samsung health AI models analyse wearable biosignal data appeared first on AI News .

Google AI health coach to use Abbott glucose data

Industry · United Kingdom · AI News

Abbott and Google are linking continuous glucose monitoring data with Google’s AI-powered health coaching tools, giving the Gemini-powered service access to another source of personal health information. Under a multiyear agreement, data from Abbott’s Lingo continuous glucose monitor will be integrated into the Google Health app. Users will be able to view glucose trends alongside […] The post Google AI health coach to use Abbott glucose data appeared first on AI News .

Apple trained its own AI model for China with help from Alibaba

Products · United States · The Verge AI

Apple has reportedly developed a custom large language model specifically for the Chinese market in collaboration with Alibaba, marking an unusual partnership between a U.S. tech giant and a Chinese firm amid escalating geopolitical tensions. The model, trained with Alibaba’s support, is tailored to comply with local regulations and language nuances, enabling Apple to enhance its AI-powered services like Siri and on-device intelligence within China’s restrictive digital ecosystem. This move reflects Apple’s strategy to maintain competitiveness in China without relying on foreign AI models that may face regulatory hurdles.

What this means for you: Professionals and businesses should monitor how localized AI partnerships like this one may reshape global tech supply chains and compliance requirements in regulated markets.

Z.ai Ships GLM-5.3 Without Retraining the Base Model: Better at Complex Coding and Long-Horizon Tasks

Research · Global · MarkTechPost

Z.ai released GLM-5.3 on August 14, 2026. The model reuses the 743B GLM-5.2 base unchanged. Every reported gain comes from scaled post-training: more long-horizon task environments, more environment types, longer training. Terminal-Bench 3.0 moves from 4.6 to 28.3, and DeepSWE v1.1 from 46.2 to 66.9. Cybersecurity moved further than Z.ai says it planned, with CyberGym at 84.5% and ExploitBench more than doubling to 54.4%. Weights arrive in about two weeks. The post Z.ai Ships GLM-5.3 Without Retraining the Base Model: Better at Complex Coding and Long-Horizon Tasks appeared first on MarkTechPo

Meet Needle 2: An Open 45M-Parameter Tool-Calling Model That Ships as a 14MB Binary and Runs a Full Session in 28MB of RAM

Research · Global · MarkTechPost

Cactus Compute released Needle 2, an open 45M-parameter model for tool calling, device use, and structured extraction. The full model is a single 14MB binary that runs a session in about 28MB of RAM. It leads both Seal-Tools splits while targeting hardware with no GPU and no NPU. The post Meet Needle 2: An Open 45M-Parameter Tool-Calling Model That Ships as a 14MB Binary and Runs a Full Session in 28MB of RAM appeared first on MarkTechPost .

Create a Reasoning-Focused LLM: A Practical Guide to Streaming, Curating, and Fine-Tuning the SupraLabs Reasoning Corpus

Research · Global · MarkTechPost

This tutorial provides a complete workflow for building a compact, reasoning-focused language model. By streaming the SupraLabs reasoning corpus from Hugging Face, we apply quality filters and curate data for Supervised Fine-Tuning (SFT). Using SmolLM2-135M-Instruct and LoRA, we demonstrate an end-to-end pipeline—from dataset analysis and heuristic cleaning to efficient training and inference—enabling the development of specialized small models without excessive resource requirements The post Create a Reasoning-Focused LLM: A Practical Guide to Streaming, Curating, and Fine-Tuning the SupraLab

State of Open Models: Summer 2026 Observations

Models · Global · Hugging Face

State of Open Models: Summer 2026 Observations