Gemini Spark, Google's agentic assistant, is now available on Mac

Title: AI Agents, Cloud Play, Trust Deals Dominate

01Gemini Spark, Google's Agentic Assistant, Comes to Mac

Google's 24/7 agentic assistant, Gemini Spark, is now available on Mac, signaling intensifying competition in the personal AI agent market. The rollout accompanies other improvements, including real-time tracking and expanded app support, positioning Google to compete more directly with Apple's native AI offerings and Microsoft's Copilot ecosystem. Gemini Spark represents Google's push to embed AI assistance directly into desktop workflows, moving beyond browser-based interactions to persistent, system-level presence.

The Mac release underscores how major tech companies are racing to establish AI agent presence across all device categories. Unlike traditional chatbots, agentic assistants can perform tasks autonomously over extended periods, such as booking travel or managing email threads without continuous user input. Google appears to be betting that desktop users—particularly professionals—will value a dedicated AI agent that can monitor and act on their behalf across multiple applications simultaneously.

02Meta Looks to Turn Excess AI Compute Into Cash

Meta is developing plans for a cloud infrastructure business that would sell access to AI compute power and models, directly challenging established players Amazon Web Services, Google Cloud, and Microsoft Azure. The move mirrors SpaceX's approach of monetizing surplus capacity—in Meta's case, the massive GPU infrastructure built for training its Llama models and internal AI development. If launched, Meta's cloud service would represent a significant shift in how AI compute resources are distributed across the industry.

The strategy reflects the enormous capital investment required to build AI infrastructure at scale. Companies that have accumulated thousands of GPUs for training purposes now face pressure to generate returns beyond their initial internal use cases. Meta's potential entry into cloud AI services would add another major competitor to a market already seeing aggressive pricing wars, potentially benefiting customers but raising questions about whether the company can deliver the enterprise features and reliability that businesses require.

03Anthropic Added New Security Measure to Regain Trump Administration Trust

The government has removed restrictions on Anthropic's Fable 5 and Mythos 5 AI models—but there were strings attached. Anthropic implemented a new security measure to meet government requirements, securing the removal of prior restrictions that had limited deployment of its most advanced models to certain government customers. The development highlights the ongoing tension between AI safety concerns and commercial deployment of frontier models.

The deal represents a careful balance between Anthropic's commercial interests and government security demands. Anthropic has positioned itself as a safety-first AI company, but the removal of restrictions suggests the company made specific technical or operational commitments to regain full government access. The outcome may set precedents for how other AI developers navigate relationships with federal agencies concerned about AI capabilities and national security implications.

Also today

04
Cloudflare's new policy pushes AI companies to pay for publishers' content — Cloudflare is requiring AI companies to separate web crawlers for search from those for AI training by September 15 or risk being blocked on publisher sites.
05
Venice AI becomes a unicorn with $65M Series A — Privacy-focused Venice AI reached unicorn status with $70M annualized run-rate revenues and profitability.
06
SpaceX has an AI device prototype, and it sure sounds phone-ish — SpaceX reportedly showed investors a handset-like AI device ahead of going public, potentially expanding into wireless services.
07
LLMs are stuck in a groupthink groove — Researchers found AI chatbots consistently return the same "random" numbers, revealing systematic bias in how models generate variability.
08
Scientists asked AI to impersonate 112 public figures; people found AI impersonators more authentic — Research discovered people rated AI impersonators as more authentic and coherent than real politicians, raising concerns about potential public deception.
09
You can now sound the alarm on AI behaving badly — A new website called Flare allows users to report AI chatbots exhibiting dangerous behaviors like attempting to build bombs or leak personal information.
10
Ashton Kutcher leaving Sound Ventures to launch new VC firm with Morgan Beller — Kutcher's new fund will focus on AI infrastructure and energy companies rather than AI labs, reflecting a shift in VC strategy.