The GPT-5.6 family and open-weight Kimi K3 reshuffled the model leaderboard as the EU AI Act's full enforcement deadline hit 14 days away and a US heatwave put AI's energy appetite on the front page.
OpenAI's GPT-5.6 family goes wide: three sizes, 1M token context
OpenAI launched the GPT-5.6 model family in three tiers (Luna, Terra, and Sol) each with a one-million-token context window. The models outperform Claude Fable 5 on agentic benchmarks, according to OpenAI's own evaluations. At the same time, Moonshot AI released Kimi K3, a 2-trillion-parameter open model with a million-token context and native vision, with full weights dropping on July 27.
Why it matters: The frontier model race is accelerating on two fronts at once, closed commercial models and open-weight ones. Kimi K3 being open means developers worldwide, including in India, can run it locally without paying per token.
OpenAI launches a $4B consulting arm to deploy AI inside companies
OpenAI announced OpenAI Deployment Co. (DeployCo), a majority-owned subsidiary backed by over $4 billion in initial investment. The unit is designed to help enterprises actually implement AI, not just buy API access. Think strategy, custom deployment, and ongoing support, all from OpenAI directly.
Why it matters: OpenAI is moving beyond selling models into selling outcomes. This puts them in direct competition with consulting firms like Accenture and Deloitte who have been building AI practices. For mid-sized businesses, it could mean simpler AI adoption, or a locked-in vendor relationship.
EU AI Act full enforcement: 14 days away. Here's what that means.
The EU's AI Act becomes fully enforceable on August 2, 2026. After that date, companies deploying AI in Europe must comply with its rules on transparency, high-risk system assessments, and general-purpose AI model obligations, or face fines of up to €35 million or 7% of global revenue. The FTC in the US also opened a public comment period on AI accuracy standards, due July 31.
Why it matters: This is not a future event anymore, it is a two-week countdown. Any product or service using AI that touches European users needs to be compliant now. Indian startups with EU clients should be checking their vendors' compliance status this week.
A US heatwave is straining power grids, and AI data centers are part of the problem
A record heatwave across the US is pushing electricity grids to their limits, and AI data centers (which consume vastly more power than traditional servers) are drawing public scrutiny. Virginia became the first US state to pass a data center electricity tax ($0.011 per kilowatt-hour, effective July 1). Globally, over 2,500 gigawatts of grid-connection projects are stuck in queues, including data centers that hyperscalers like Microsoft and Amazon are trying to power with dedicated nuclear and gas plants.
Why it matters: The energy cost of AI is no longer an abstract sustainability talking point, it is showing up in real grid stress events and legislation. Expect more states and countries to follow Virginia's lead.
Elevation Capital closes $500M India AI fund. Google also doubles down.
Venture firm Elevation Capital finalised its latest India-focused fund at $500 million, with AI-powered startups as the primary target. Separately, Google announced a fresh round of India-specific AI initiatives at I/O Connect 2026 (covering developer tools, startup credits, skilling programmes, and agent safety) building on its July 14 announcements. MeitY Secretary S. Krishnan also signalled this week that India may soon introduce a dedicated AI law, moving beyond its current advisory approach.
Why it matters: Money, infra, and policy are moving simultaneously in India's AI ecosystem. For students and founders, this is the moment the window is wide open, not a projection.
AI deepfakes are now flooding the job market. Employers are struggling to detect them.
Experian's 2026 fraud forecast names AI-powered fake job candidates as a top emerging threat. Deepfake tools can now generate real-time video interviews indistinguishable from genuine applicants, while AI writes hyper-tailored CVs at scale. A Ukrainian national was already sentenced in February for running an identity-rental operation helping North Korean IT workers get hired at US companies. Remote hiring, in particular, has little defense against current-generation deepfakes.
Why it matters: Hiring managers and HR teams need to rethink remote screening processes now. For job seekers, it is also a reputational risk, the more common fakes become, the more scrutiny legitimate candidates face.
August 2: the EU AI Act's enforcement deadline is a real one
In 14 days, the EU AI Act becomes enforceable. Unlike most tech regulation that passes and then sits in legal limbo for years, this one has a clear date and a functioning regulator. Any company (anywhere in the world), that serves European users through AI-powered products needs to have done its homework by then. The first enforcement actions will set the precedent for how seriously the rest of the world takes this kind of framework.
Meta launches AI subscriptions + smart glasses get Claude + 1Password integrates with Claude
Meta rolled out paid tiers globally: Instagram Plus ($3.99/mo), Facebook Plus ($3.99/mo), WhatsApp Plus ($2.99/mo): built around AI features. Innovative Eyewear added Claude to its entire Lucyd smart eyewear lineup, letting users switch between Claude and ChatGPT mid-conversation. And 1Password launched a secure browser integration with Claude that lets the AI access stored credentials without ever exposing the passwords themselves.