Models & Products 29 September 2026
OpenAI releases GPT-6.1 Sol at one-fifth the price of its top model
At its DevDay event, OpenAI released GPT-6.1 Sol. It costs $2 per million input tokens and $10 per million output tokens, one-fifth of the price of GPT-6 Astra. OpenAI says it performs close to Astra on coding, computer use and office work. It is available in ChatGPT Work and Codex for paid plans and through the API, but not yet in the main ChatGPT chat.
Why it matters: Indian software firms and startups that build on AI can get near-top quality for much less money. This makes AI tools cheaper to offer to small businesses.
Models & Products 29 September 2026
OpenAI launches 'dots', agents that keep working on their own
OpenAI launched dots, AI agents inside ChatGPT that each run on their own computer in the cloud. A user gives a goal, and the agent keeps working on it even when the user is away. It can connect to more than 4,000 apps and can be reached in ChatGPT, Slack and Teams. The first dot comes with Pro and Business Premium plans. Enterprise, Edu and Healthcare customers get a beta. OpenAI says actions that touch a user's accounts are checked automatically, and that agents can still make mistakes.
Why it matters: Office tasks such as research, follow-ups and reports can be handed to software that works in the background. Businesses should decide in advance which accounts and data such an agent may use, and should check its work.
Safety & Security 30 September 2026
Google gives Gemini 4 Argon to cyber defenders first
Google announced Gemini 4 Argon, its most advanced model. It is first going to trusted cyber defenders in its Fairwind Program, which has more than 650 partners, including governments. Google says the model can find, confirm and fix serious software flaws on its own, and that it found a flaw in widely used hospital software that earlier models missed. Google has not given a date or a price for wider release.
Why it matters: Releasing the model to defenders first is meant to let them fix flaws before attackers can use the same tool. Indian banks, hospitals and government offices depend on software that such models can now check.
Safety & Security 1 October 2026
Microsoft report: attackers turn new flaws into attacks in under a day
Microsoft released its 2026 Digital Defense Report, covering July 2025 to June 2026. It says the median time from finding a software flaw to using it in attacks has fallen well below 24 hours. Companies typically take 30 to 60 days to fix serious flaws. It says AI is now used for phishing, finding flaws and writing malware. Nearly 40,000 flaws were published in the first half of 2026 alone.
Why it matters: Small businesses in India need to install software updates quickly, use strong logins and two-step verification, and be careful with emails. Waiting weeks to fix systems is now much riskier.
Policy & Rules 29 September 2026
Top US AI companies sign a voluntary safety pledge at the White House
The heads of Anthropic, OpenAI, Google, Meta, xAI and Nvidia met at the White House and signed a 'Joint Commitment on Frontier Responsibilities'. They promised internal checks for cyber, biological and chemical risks, independent outside auditors, and board-level oversight. They also agreed to meet regularly on shared safety standards. Critics say the pledge is voluntary and is meant to head off government rules.
Why it matters: The pledge is not law, so it gives users in India and elsewhere no legal protection. It does show which safety steps the companies say they will take.
Policy & Rules 30 September 2026
US FTC investigates OpenAI and Anthropic over AI safety
The US Federal Trade Commission confirmed it is investigating OpenAI and Anthropic. It is looking at possible unfair or deceptive practices and harm to consumers. The agency may issue formal demands for information. It can enforce consumer protection law, which can lead to civil penalties.
Why it matters: Indian businesses use these companies' products. Any safety changes forced by the US regulator are likely to apply to their services worldwide.
Policy & Rules 30 September 2026
California law says AI alone cannot fire or discipline workers
California Governor Gavin Newsom signed SB 947, the 'No Robo Bosses Act', along with other AI bills. When an employer relies mainly on an AI system to fire or discipline someone, a human must review and confirm the decision. The worker must also get a written notice. The law takes effect on 1 July 2027.
Why it matters: Indian IT and HR-software firms that serve US clients will need to build human review and notices into their products. The law may also serve as a model for rules elsewhere.
Business & Funding 28 September 2026
Anthropic's IPO filing shows $518 billion in infrastructure commitments
Details of Anthropic's IPO filing were reported on 28 September. The filing was made confidentially in June. It shows about $4.59 billion in revenue for 2025 and $518 billion in commitments for cloud, computing and other infrastructure, about 80% of which cannot be cancelled. The company is aiming for a valuation above $2 trillion, and the listing is expected before late November.
Why it matters: The numbers show how much AI companies are spending on computing compared with what they earn. That spending affects how fast AI prices fall for users, including in India.
India 26 September 2026
Google tests buying Flipkart products inside Gemini in India
Google began testing a 'Buy' button for some Flipkart products in Gemini and in AI Mode in Search in India. Shoppers can check out through a Flipkart-branded flow without leaving the AI chat. The test covers smartphones, electronics and accessories for a small group of users. Google plans a wider rollout later in October, before the festive season.
Why it matters: Indian shoppers may soon buy through AI chats instead of browsing websites. Sellers may need to make sure their product listings are clear enough for AI tools to show them.
India 30 September 2026
Kerala opens IndiaAI Centre of Excellence with 15 startups
Kerala's IT and AI Minister P.K. Kunhalikutty launched a Centre of Excellence in AI at Technopark, Thiruvananthapuram, set up under MeitY's IndiaAI Mission. Digital University Kerala runs it. Its first group has 15 startups working in healthcare, farming and climate, legal tech and industrial AI. They will get cloud credits, mentoring and tools. The plan also includes certification courses for 1,600 students over four years.
Why it matters: It gives AI startups outside the big metros access to support under the national IndiaAI Mission. Some of the startups work on legal and document tools that small businesses and citizens could use.
Compiled from public reports and official announcements. Follow the source links for full details. Next update: Saturday, 17 October 2026, 2:00 PM IST.