26 Aug 2025

AI News Digest

🤖 AI-curated 6 stories

Today's Summary

NVIDIA’s latest move with the Jetson Thor is like handing robots a supercharged brain, letting them handle complex AI tasks right on the edge without always phoning home to the cloud. Meanwhile, Perplexity’s new subscription model is trying to make peace with publishers by sharing the wealth from AI-driven content, possibly nudging other AI players to rethink how they treat journalistic work. On a different note, TikTok’s shift towards AI moderation is part of a wider trend, but it’s stirring up concerns about job security and the ever-evolving balance between human oversight and AI efficiency.

Stories

NVIDIA ships Jetson Thor — a Blackwell‑powered 'robot brain' for next‑gen humanoids and edge robots

NVIDIA announced general availability of Jetson AGX Thor (developer kit) and Jetson T5000 production modules, a Blackwell‑GPU powered robotics computer that delivers up to ~2,070 FP4 TFLOPS and 128 GB memory for on‑device generative reasoning. The launch (developer kit available now) targets humanoid and industrial robots by enabling multiple generative AI models and low‑latency multimodal sensor processing at the edge — a step toward running sophisticated vision‑language‑action and agentic models locally rather than in the cloud. Impact: Jetson Thor accelerates the 'physical AI' market (robotics, logistics, healthcare, agriculture), lowers cloud dependency for robotics vendors, and tightens NVIDIA’s position as the dominant supplier of hardware for high‑performance on‑device AI. ([investor.nvidia.com](https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Blackwell-Powered-Jetson-Thor-Now-Available-Accelerating-the-Age-of-General-Robotics/default.aspx))
Read more → NVIDIA (press release)

Perplexity launches Comet Plus and a $42.5M revenue‑share program to pay publishers

Perplexity announced Comet Plus — a $5/month tier for its Comet AI browser — and a publisher revenue‑sharing initiative that initially allocates about $42.5 million to participating outlets. Under the plan Perplexity says ~80% of Comet Plus revenue (minus compute) will go to publishers based on human visits, in‑response citations, and agent actions. Why it matters: the move is an attempt to address publisher backlash and lawsuits over AI use of journalistic content by creating a subscription‑funded compensation model; it could pressure other AI search and agent companies to adopt similar publisher payment schemes and reshape the economics of AI‑driven search/agents. ([indianexpress.com](https://indianexpress.com/article/technology/artificial-intelligence/perplexity-new-comet-plus-subscription-revenue-sharing-publishers-10212605/))
Read more → The Indian Express

TCS names Amit Kapur to lead new AI & services transformation unit as Indian IT doubles down on AI

Tata Consultancy Services has appointed Amit Kapur to head a newly formed AI and services transformation unit, effective September — a strategic leadership move as the Indian IT giant scales AI offerings while recently announcing large job cuts. The change signals how legacy outsourcers are reorganizing around AI products and services to capture subdued tech spending, and could presage further industry restructuring and workforce reshuffles across the $283B Indian outsourcing sector. ([reuters.com](https://www.reuters.com/world/india/indias-tcs-names-amit-kapur-head-newly-formed-ai-unit-company-memo-shows-2025-08-26/))
Read more → Reuters

TikTok’s trust-and-safety shake-up: hundreds of moderator roles put at risk as AI takes over moderation

The Guardian reports TikTok has put several hundred UK trust-and-safety/moderation roles at risk as part of a global reorganisation that shifts more content-review work to AI systems. The move — happening amid fresh regulatory scrutiny (the Online Safety Act) and union pushback — underscores a broader industry trend: platforms are increasingly automating moderation, raising questions about safety, offshoring, and the future of human moderator jobs. ([theguardian.com](https://www.theguardian.com/technology/2025/aug/25/artificial-intelligence-openai-tiktok))
Read more → The Guardian

Google expands AI Mode globally and adds agentic features for reservations and personalization

Google rolled out AI Mode to 180 more countries (English) and introduced new agentic capabilities that let the feature take multi-step actions — starting with restaurant reservations — plus personalized search results based on user preferences. The move brings more agent-like automation into Search (via Google Labs) and couples generative assistance with real‑world actions (booking, ticketing, appointments) for paid AI Ultra users. Why it matters: this makes powerful, agentic AI accessible to many more users and signals a shift from static answers toward actionable AI features embedded in everyday apps — raising competition with other assistant ecosystems and offering new workflows for productivity and consumer services.
Read more → TechCrunch

GitHub launches an 'agents panel' — a mission control to delegate and manage Copilot coding agents

GitHub introduced an Agents panel (a popup 'mission control') that allows paid Copilot subscribers to assign tasks to Copilot coding agents from any page on github.com, track progress, and review agent-created pull requests without leaving their workflow. The feature aims to make AI agents act like background teammates — drafting changes, running tests and preparing PRs autonomously. Why it matters: it simplifies using AI agents for real coding work, lowers friction for adopting autonomous developer tooling, and expands practical 'vibe‑coding' workflows for teams and individual developers while raising questions about governance, review practices and security in agentic development.
Read more → ITPro