OpenAI's GPT-6.1 Sol landed in ChatGPT Work, New York banned deceptive AI chatbots, and the global deepfake crackdown entered its final countdown.
OpenAI's GPT-6.1 Sol is now live in ChatGPT Work and Codex
OpenAI rolled out GPT-6.1 Sol to ChatGPT's Work interface and Codex today, available to Plus, Pro, Business, Enterprise, and Education subscribers. It is not yet available in regular Chat. The company has not published a full technical brief, but early users are noting faster, more precise outputs in coding and document tasks.
Why it matters: If you use ChatGPT for work tasks, you have a new model option today. The tiered rollout, work tools first, also signals OpenAI's priority: enterprise and professional users before the general public.
Google DeepMind's Genie 4 can build an interactive world from a single image
Google DeepMind launched a limited research preview of Genie 4, a model that generates real-time interactive environments from an image or text prompt. Access is restricted to trusted testers, Google AI Ultra subscribers in Project Genie, and Gemini API early partners. Think of it as AI that doesn't just describe a scene but lets you explore it.
Why it matters: This points toward AI-generated games, training simulations, and interactive learning tools built in minutes. It is in early access now, but the applications for education and professional training are worth watching closely.
New York just banned AI chatbots from pretending to be human
Governor Hochul signed two AI bills into law on October 28. The first bars AI chatbots from claiming to be real people or expressing emotions they do not have. The second restricts companies from offering certain chatbot features, including companion-style interactions, to users under 18 or users whose age has not been verified. New York is the first US state to make this a legal requirement.
Why it matters: This is the first US state law targeting deceptive AI identity directly. It sets a template other states are likely to follow, and it puts real pressure on the companion-AI and edtech sectors to build age verification before they ship.
Indic AI startups have raised $399M in 2026. One company got 87% of it.
A new report tracking Indic-language AI startup funding puts the 2026 total at about $399 million across eight rounds. Sarvam AI accounts for $346 million of that through two rounds. The remaining seven startups split just $54 million between them. Investors are clearly concentrated in one flagship bet while the broader ecosystem remains early-stage.
Why it matters: The concentration is both a sign of confidence in India's AI potential and a gap in how that capital is spreading. For founders building outside of language AI, there is still significant white space and arguably less competition for early funding.
The EU's AI deepfake ban takes effect December 2. Platforms have five weeks.
The EU AI Omnibus amendments adopted on October 13 include a ban on AI systems that generate non-consensual intimate imagery. The ban becomes enforceable on December 2, 2026. Fines for violations can reach 35 million euros or 7% of global annual turnover, whichever is higher. Platforms operating in Europe need to have removal systems in place before that date.
Why it matters: This is one of the most specific and time-bound enforcement actions to come out of the EU AI Act so far. The five-week countdown is real, and the fine structure puts serious financial risk on platforms that ignore it.
Meta's Muse Wonder models are available now on the Meta Model API
Meta launched its Muse Wonder model family today on the Meta Model API and inside Muse Code, its developer environment. The fuller Muse Wonder version is limited to Muse Code subscribers. The models are positioned for creative and generative tasks, though Meta has not published benchmarks.
Why it matters: Meta is building a developer ecosystem around its own models, not just distributing through others. If you are building AI-powered creative tools, this is a new API option worth evaluating alongside the usual choices.
The global deepfake crackdown is no longer a future event
This week alone: New York signed laws against deceptive chatbots, the EU set a December 2 deadline for its deepfake ban, Queensland introduced criminal penalties for non-consensual deepfakes, and Pakistan opened public consultation on election deepfake rules. India's IT Rule amendments with mandatory synthetic content labelling are also in motion. These are not proposals. They are laws on a countdown. The question for platforms, creators, and anyone who uses AI-generated content is whether their current practices will be legal in 60 days.
Akamai announced a new sandbox environment that runs agent-written code in a restricted container, a direct response to the security risks of autonomous AI agents executing arbitrary code. Separately, Mistral released four open-weight probabilistic time-series forecasting models on Hugging Face, free for anyone building prediction or analytics tools.