GT Protocol AI Digest №59: Sora, Sonnet, and Stargate
Intro
This past week has brought a wave of significant AI developments — ranging from Hollywood’s uneasy experiments with generative storytelling to breakthroughs in quantum chips, infrastructure partnerships, and the ongoing arms race in productivity AI. At the same time, OpenAI’s latest video tools and Microsoft’s human-like Copilot avatars are blurring the line between reality and synthetic media, raising fresh questions about creativity, trust, and safety. Let’s dive into the 13 key stories shaping the future of AI.
Special Topic: Chips & computing breakthroughs
• Quantum chips enter “real-world ready” phase: New research published in Science shows quantum processors demonstrating robustness, error correction, and scalability that push them beyond lab demos into potential industrial use cases. These advancements signal that quantum chips are moving closer to integration in practical computing environments, where they could eventually complement AI workloads. Read more here
1. OpenAI in the limelights
• Sora 2 debuts: OpenAI has announced Sora 2, an upgraded video generation model with better realism, support for personal cameo-style uploads, and improved scene coherence. Initially available in the U.S. and Canada, it launches with a free tier and optional Pro plan. The release demonstrates OpenAI’s commitment to making high-quality video synthesis more accessible, even as concerns about misuse persist. Read more here
• Parental controls roll out for ChatGPT: OpenAI has introduced parental controls to its platform, giving guardians the ability to manage teen accounts. Features include linking accounts, disabling image generation, enforcing “quiet hours,” and restricting memory functions. While parents cannot view conversations directly, OpenAI promises alerts in cases of severe safety risks. The rollout starts on the web, with mobile support coming soon. Read more here
• OpenAI launches a social video app with deepfake features: OpenAI has released a new iOS social video app powered by Sora that lets users generate hyper-realistic clips, including swapping faces to mimic friends. The app blurs the line between entertainment and impersonation, raising major questions around digital consent, abuse prevention, and platform responsibility, even as it offers new forms of playful creativity. Read more here
• “TikTok for deepfakes” critique: A Verge review warns that OpenAI’s video app is essentially becoming a TikTok-style platform for synthetic media, with content that is increasingly indistinguishable from real footage. Analysts fear this leap in realism and accessibility will worsen misinformation challenges, deepen trust issues online, and make verification systems critical for the future of media consumption. Read more here
2. Microsoft’s Updates
• Copilot gets human-like faces (in Labs): Microsoft is piloting Copilot avatars — 40 stylized digital faces that users can chat with. This experiment aims to make AI more relatable and engaging, but also tests ethical boundaries around anthropomorphism, emotional attachment, and user trust in human-like digital assistants. Read more here
• Microsoft’s “vibe working” lands in Excel & Word: Microsoft has introduced Agent Mode in Excel and Word, allowing tasks to be broken into smaller, auditable steps, increasing transparency. The new Office Agent in Copilot chat can generate entire presentations or documents from prompts, powered by Anthropic’s models. Early benchmarking like SpreadsheetBench shows higher accuracy in stepwise reasoning, reflecting Microsoft’s push for trustworthy AI in business software. Read more here
• Microsoft 365 Premium bundles Office + AI at $19.99/mo: The company has launched a new premium subscription tier that merges Copilot Pro with Microsoft 365 Family. For the same price as ChatGPT Plus, subscribers get six Office desktop seats, 1TB of OneDrive storage per person, and advanced AI usage limits. The move is a direct competitive strike at standalone AI subscription models. Read more here
3. AI infrastructure & New Claude
• Fusion for AI datacenters? With AI driving unprecedented energy demands — U.S. datacenters already consume 4.4% of national electricity and could reach up to 12% by 2028 — companies are exploring nuclear fusion as a future power source. Microsoft has signed a 2028 power purchase agreement with Helion Energy, while Google has partnered with Commonwealth Fusion Systems. These moves show how the AI boom is accelerating long-term bets on next-generation energy. Read more here
• Samsung & SK join OpenAI’s “Stargate” initiative: Samsung and SK have committed to supporting OpenAI’s Stargate project, an ambitious plan to scale global AI infrastructure. With targets like 900,000 DRAM wafer starts per month and collaborative data center development in Korea, the initiative aims to position the country as a top-three AI hub worldwide while strengthening supply chain resilience. Read more here
• Anthropic’s Claude Sonnet 4.5 goes long-running: Anthropic has unveiled Claude Sonnet 4.5, capable of running autonomously for 30+ hours on complex coding projects, such as producing 11,000 lines of code for a chat application. The update highlights advancements in multi-agent coordination, memory, and virtual machine integration, underscoring Claude’s growing strength in handling sophisticated programming and agentic workflows. Read more here
4. GenAI & Entertainment
• Hollywood’s AI push meets reality: Generative AI boosters are working hard to get into Hollywood, pitching tools to cut costs, accelerate production, and even replace parts of the creative process. However, industry insiders point out that the technology remains expensive, often clunky, and struggles to meet professional standards in narrative and visual quality. Even high-profile experiments like OpenAI’s Critterz, rumored to cost around $30 million, highlight how far AI still has to go before it can reshape film and television. Read more here
• Australia’s culture debate intensifies: Musician and advocate Briggs has warned that Australia risks abandoning its artists if it fails to enact copyright protections against AI exploitation. He argues that once AI systems absorb creative works without consent, it’s impossible to “get the genie back in the bottle.” His comments add urgency to ongoing debates about how AI training intersects with intellectual property, artist livelihoods, and national cultural policy. Read more here
Outro
AI is rapidly evolving from experimental demos into mainstream infrastructure — touching entertainment, productivity, energy, and everyday communication. Yet as models become more powerful and human-like, the balance between innovation and responsibility grows more fragile. For GT Protocol, these shifts are more than headlines — they are signals of where AI’s influence is heading, and how it will reshape industries, finance, and the digital economy. Stay tuned, stay critical, and stay ahead.
