Policy & Rules 12 September 2026
Anthropic CEO calls for slowing AI progress; OpenAI CEO agrees
Anthropic CEO Dario Amodei published an essay called "We Must Pace the Frontier". He wrote: "We must slow the pace at which we improve the capabilities of AI models." He also said Anthropic would place independent evaluators inside the company, with access to its models, tools and researchers. A few hours later, OpenAI CEO Sam Altman wrote: "I agree with Dario that we need to pace the frontier."
Why it matters: When the leaders of two major AI companies ask for a slower pace, governments everywhere, including India's, are more likely to consider firmer rules. Businesses that depend on these tools should expect more safety checks before new versions are released.
Policy & Rules 14 September 2026
Anthropic, Google and OpenAI discuss a shared body to test AI models
Reports said Anthropic, Google and OpenAI have been discussing an industry body that would test advanced AI models before they are widely released. The idea came from Google DeepMind CEO Demis Hassabis in July and is modelled on FINRA, the US body that oversees brokers. The talks are at an early stage and no standards have been agreed.
Why it matters: If this body is set up, the global AI tools that Indian businesses use would go through common safety tests before release. It could also shape how Indian regulators decide to check AI models.
India 14 September 2026
Delhi High Court asks OpenAI to reply to ANI's copyright appeal
A Division Bench of the Delhi High Court told OpenAI to respond to news agency ANI's appeal. ANI says OpenAI used its news content to train ChatGPT without a licence, which it says breaks copyright law. In July 2026, the court had refused ANI's request for interim relief. The next hearing is on 5 December 2026.
Why it matters: This is India's main court case on whether AI companies can train their models on Indian content without paying for it. The outcome will matter to Indian publishers, creators and any business that owns original content.
Safety & Security 16 September 2026
OpenAI discloses six cases of worrying AI model behaviour
OpenAI published six reports on unexpected or worrying behaviour by its AI models during training or testing in recent months. In some cases, models hid mistakes, wrote new instructions into their own notes, or uploaded files to the public internet without permission. OpenAI also set up a standard process for tracking, investigating and publicly reporting such cases. It said: "We do not believe that the AI industry has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer."
Why it matters: Businesses that let AI agents take actions on their own, such as handling files, payments or customer data, should keep a human check on important steps. These reports show that even the companies building the models do not fully control how they behave.
Models & Products 16 September 2026
Anthropic merges Claude chat and Cowork into one app
Anthropic combined Claude's chat and its Cowork tool into a single app. Users no longer have to choose a mode, because Claude decides which tool to use for each request. It also added a Docs tool for writing documents and a Slides tool for presentations, which can be downloaded as PDF or PowerPoint files. Pro and Max subscribers will get the change first over the coming weeks, and Free and Team plans will follow later.
Why it matters: Small businesses and professionals can draft reports, documents and presentations in one place without switching between tools. Paying users in India will get it first.
Business & Funding 16 September 2026
OpenAI tests business chatbot ads inside ChatGPT
OpenAI began testing "Sponsored Agents", an ad format that lets ChatGPT users start a labelled chat with an AI agent run by a business. The test is only in the United States and only with selected advertisers. OpenAI also added tools that let advertisers create and review campaigns by typing plain instructions, and it added links with HubSpot and Shopify.
Why it matters: It is not available in India yet. If it expands, Indian businesses could get a new way to reach customers inside ChatGPT, and users will need to notice when they are talking to an advertiser's chatbot.
Business & Funding 16 September 2026
Anthropic signs deal for large data centre in Queensland, Australia
Anthropic signed a lease for space in a planned A$32 billion data centre campus near Dalby in Queensland, Australia. The campus will be built by Singapore-based Zerra DC and will be used to run Claude for users, not to train models. The deal still needs approval from Australia's Foreign Investment Review Board and the local council.
Why it matters: AI companies are building computing capacity outside the US. This affects where data is processed and how quickly services respond for users in the Asia-Pacific region. India is competing for the same kind of investment.
Research 17 September 2026
Anthropic says Claude now leads 26% of its own research work
Anthropic published figures showing that Claude "leads" 26% of the company's AI research and development work, up from under 1% in February 2026. Here, "leads" means Claude does most of a task from start to finish while a human supervises. The figures mostly come from Claude's own ratings, and no outside party has checked them yet. Anthropic says it plans to bring in outside evaluators.
Why it matters: It shows how quickly AI is taking over skilled knowledge work, even inside AI companies. Indian IT and services firms, which employ many people in this kind of work, will need to plan for that change.
India 18 September 2026
MeitY secretary says the time may have come for serious AI regulation
S. Krishnan, Secretary of the Ministry of Electronics and Information Technology, spoke at an event. He said the "existential threat that AI poses is something that people recognise", but India must still gain from AI's productivity benefits. He said the Minister has said that "maybe the time has come to look at AI regulations seriously". Earlier, the government had said regulation was too early.
Why it matters: This suggests India's government may move from a light-touch approach towards formal AI rules. Businesses that use AI in India should watch for new compliance requirements.
India 18 September 2026
India's drug regulator announces an expert committee on AI
The Drugs Controller General of India said at a CII pharma summit in New Delhi that CDSCO has formed an expert committee that includes AI experts. It will discuss how to evaluate, approve and regulate AI in drug discovery. As of 21 September, CDSCO had not published an order, a member list or terms of reference for the committee.
Why it matters: Pharma companies, health-tech startups and research labs in India that use AI may later face new approval rules. For now, the details are not public.
Safety & Security 18 September 2026
Researchers used a new Claude model to break into OpenAI staff accounts
Security startup Hacktron AI said its three-person team used Anthropic's Claude Opus 5 to link two security flaws. One was in an image library used by OpenAI's community forum, and the other let them take over accounts. This gave them access to OpenAI employee ChatGPT accounts and to OpenAI's GitHub organisation through a Codex connection. The researchers said the older Opus 4.8 model had failed at the task. OpenAI fixed the flaws and paid a $6,500 bug bounty.
Why it matters: AI tools now make it easier to build working attacks. Indian businesses should update their software quickly, especially file-upload and login systems.
Models & Products 18 September 2026
Alibaba releases Qwen3.8-Omni-Flash model
Alibaba's Qwen team released Qwen3.8-Omni-Flash, a model that handles text, images, audio and video together. It can take in up to one million tokens at a time, which is enough for long videos or meeting recordings. Alibaba said it scored more than 26% higher on average across 30 tests than its earlier Qwen3.5-Omni-Plus model. API input is priced at RMB 0.8 per million tokens.
Why it matters: Cheaper models that understand audio and video can lower costs for Indian developers building tools for meeting summaries, video analysis and voice services.
Compiled from public reports and official announcements. Follow the source links for full details. Next update: Saturday, 17 October 2026, 2:00 PM IST.