AI never sleeps, and neither do the people involved in it. Hence, as another week goes by, FE’s resident AI gossipmonger is back with another round of the AI weekly roundup, which has now started to gather a loyal set of readers. Thanks to everyone for taking out their precious time to read. Let’s get on with everything that AI has given us to talk about this week.
Kunal Shah dominated the headlines this week – the CRED guy who now will take care of your favourite messenger app from Meta. Nobody saw Kunal’s ascend to the throne, especially in the week following the ban of its biggest rival, Telegram. Telegram returned to service but it was Kunal who stole the spotlight.
Then there was Amazon’s major investment announcement as well as our dear Uncle Sam pulling the handbrake on his latest version of ChatGPT. From unprecedented White House interventions blocking frontier model deployments to massive architectural market restructuring, the AI landscape continues shifting from “cool software tricks” to high-stakes geopolitics.
Grab your coffee; it’s time to get gossiping.
Meta’s $900mn bet on Kunal Shah
Meta’s absolute blockbuster announcement sent shockwaves from Menlo Park to Bengaluru. In a stunning executive coup, Mark Zuckerberg has appointed CRED founder Kunal Shah as the Global CEO of WhatsApp, succeeding long-time head Will Cathcart. But Zuckerberg didn’t just hand over the keys to the world’s largest messaging platform. Meta simultaneously dropped a massive $900 million investment directly into CRED.
For years, WhatsApp has been looking for its true monetisation holy grail. It dominates communication globally, especially in emerging markets like India, but has largely remained an expensive utility rather than a revenue powerhouse.
Enter Kunal Shah, a man who built an empire by capturing the highest-intent, most credit-worthy top 1% premium audience in India. By merging Shah’s deep understanding of premium consumer behaviour and transactional monetisation with WhatsApp’s unparalleled global distribution, Meta is signalling the birth of the ultimate global super-app, something like WeChat.
With Nasscom predicting a massive surge in autonomous AI agents, Shah’s mandate is clear: transform WhatsApp into an ecosystem where AI agents handle everything from your premium wealth management to luxury bookings and enterprise logistics natively inside the chat window.
Uncle Sam pulls the handbrake on GPT-5.6
The US administration actively stepped in to freeze the wide-scale commercial deployment of OpenAI’s highly anticipated flagship model, GPT-5.6. The Office of the National Cyber Director and the Office of Science and Technology Policy (OSTP) cited critical national security anxieties. Specifically, federal evaluators flagged the model’s alarming proficiency in autonomously mapping and exploiting highly complex zero-day code vulnerabilities. OpenAI CEO Sam Altman has reportedly confirmed to internal teams that access during this restricted rollout phase will now require strict government clearance.
As for Anthropic…
US Govt starts unbanning Mythos
Anthropic has received authorisation to redeploy Claude Mythos 5, their most advanced cybersecurity model, specifically to select US organisations responsible for operating and defending critical infrastructure. While access is being quickly restored for these critical sectors, the model remains restricted for the broader public. Furthermore, their other major model, Fable 5, remains entirely offline as Anthropic continues negotiations with government officials to secure its release for general public use.
OpenAI cooks up “Jalapeno” to escape Nvidia
To insulate itself from both regulatory scrutiny and supply-chain vulnerabilities, OpenAI officially pulled back the curtain on its first custom application-specific integrated circuit (ASIC), Jalapeño. Developed over a highly confidential nine-month sprint alongside Broadcom and Celestica, Jalapeno is custom-tailored purely for high-velocity LLM inference workloads. It’s a loud declaration of independence, proving that the frontier AI race is shifting heavily into raw silicon ownership to survive the ongoing GPU drought.
OpenAI shares data on how people use AI
According to a recent economic research paper published by OpenAI, tools like Codex are changing how knowledge work is structured, moving away from short, self-contained prompts toward delegated, long-horizon tasks that can run independently for hours. This evolution is sparking rapid adoption across legal, finance, and human resources, where workers are increasingly using parallelised AI agents to orchestrate complex data analysis and cross-functional operations. The data suggests that when workers are given broad, low-friction access to capable AI agents, workflows naturally shift away from simple assistant queries toward highly parallelised, complex, and cross-functional task execution.
Anthropic exposes Alibaba
The geopolitical tension surrounding proprietary data boiled over into Congress this week. Anthropic filed an official complaint alleging that entities deeply tied to Chinese tech powerhouse Alibaba orchestrated a massive “model distillation” campaign. The claim states that Alibaba deployed over 25,000 fraudulent accounts to hammer the Claude model with more than 28 million interactions, systematically reverse-engineering Anthropic’s advanced multi-step reasoning pathways to train competitive native models.
The $81,000 hangover of Vibe Coding
The trendy concept of “vibe coding” hit a financial wall at fintech startup Slash. An enthusiastic employee accidentally left an unoptimised agent loop running while developing a minor casual video game, racking up a jaw-dropping $81,267 bill in raw API tokens over a single weekend. Department heads everywhere are scrambling to implement rigid spending firewalls.
Nasscom forecasts a $400 billion Agentic AI windfall
The Nasscom US CEO Forum released a highly optimistic economic brief. India’s tech services sector already draws an estimated $10 billion to $12 billion in pure-play AI revenues, backed by a formidable pool of two million upskilled professionals. The true gold rush, however, is the incoming wave of Agentic AI. Nasscom projects that the complex task of orchestrating, securing, and rebuilding broken corporate workflows for autonomous enterprise agents will unlock an additional $300 billion to $400 billion in market runway for Indian IT firms by 2030.
Amazon drops a $48 billion India bet
Fresh off his high-profile meeting with Prime Minister Narendra Modi in New Delhi, Amazon CEO Andy Jassy announced an absolute monster of an update. Amazon is stepping up its long-term commitment to India with a total proposed investment of $48 billion by 2030. A massive $21 billion of that war chest is being funnelled strictly into scaling up the country’s AI and AWS cloud infrastructure. The plan includes expanding mega data centres in Mumbai and Hyderabad to provide startups, corporations, and government organisations with localised access to custom AI chips and managed AI services.
Not stopping at enterprise, Jassy also pledged AI education tools for 4 million government school students and AI-led digitisation for 15 million small businesses.
Axis Bank hires an AI officer
Axis Bank went a step ahead of its peers and appointed Namrata Dubashi, a veteran former partner at McKinsey, as its dedicated enterprise Artificial Intelligence Officer. Dubashi’s mandate focuses heavily on institutionalising AI for advanced risk mitigation, real-time threat detection, and systemic credit underwriting. This move mirrors the broader operational scale-up of the central government’s Rs 10,300 crore IndiaAI Mission, which reported this week that nearly 89% of all newly registered domestic tech startups have actively integrated AI workflows into their baseline product architecture.