Daily AI Brief

Copyright opt-out vs opt-in: OpenAI's policy allows copyrighted material to be used by default unless rights‑holders explicitly opt-out. This approach has sparked concern, with critics warning that creators may be included inadvertently unless they act.

Likeness, consent, and "cameos": The system requires user verification (voice + image) to create a cameo. Users can revoke, restrict or control how their likeness is used.

Moderation and misuse risk: OpenAI incorporates guardrails: blocking content involving minors, restricting uploads of photorealistic persons initially, and evolving moderation thresholds.

Geographic and regulatory rollout: The app is launching only in the U.S. and Canada initially; extension to other jurisdictions will raise additional regulatory challenges—especially around defamation, impersonation, deepfakes, and content liability.

Industry pushback: Hollywood and rights‑holder groups are mobilising to demand tighter control and protections, particularly about derivative use and default inclusion of content.

Synchronized audio + video: Sora 2 is the first generation from OpenAI to generate synchronised dialogue, ambient sound, and effects in sync with visuals.

Physical realism, continuity, and steerability: The model emphasises accuracy in physics (e.g. correct bounce, inertia, object interaction) and allows more control across camera movement and scene dynamics.

Limits and uncertainty: Many output constraints remain unpublished (maximum clip length, resolution, framerate). Some edge cases still exhibit artefacts, discontinuities or unrealistic behaviour.

Comparative positioning: Analysts contrast Sora 2 with competing video models (e.g. Runway Gen‑4, Google's Veo) on criteria such as consistency, realism, and openness.

Cerebras Systems — Series G raise: The company raised US$1.1 billion at an $8.1 billion valuation, led by Fidelity, Atreides and including 1789 Capital (linked to Donald Trump Jr.).

Delay of IPO: Previously preparing to go public, the company is now postponing the IPO following this huge private injection and regulatory (CFIUS) scrutiny.

The influx of capital is intended to scale manufacturing, expand data centre reach, and fortify its position against incumbents.

Sora / Sora 2 app: The new social video app allows users to generate, remix, and share short AI videos. It functions in a feed model and supports cameo insertion of users' likeness.

Access & tiers: Invite-only access on iOS (U.S./Canada). Future web access via sora.com and a planned API rollout. ChatGPT Pro users may receive access to a higher-quality "Pro" variant.

System card & safety design: The Sora 2 System Card outlines capabilities, risks, and mitigation strategies (e.g. restricting uploads of realistic human video initially, watermarking, evolving moderation).

Remix & feed mechanics: The app design encourages user engagement via remixing of clips, "For You" style feed, and social interactions constrained by user consent.

Sora 2 is here — OpenAI's launch page with model details & roadmap

OpenAI is launching the Sora app, its own TikTok competitor — TechCrunch overview

OpenAI Sora 2 is Here: Key Takeaways — summary of major features and rollout

Editing vs Generating in 2025 — comparative analysis and guidance

Cerebras Systems Raises $1.1 Billion Series G — the funding announcement