One-Line Summary

OpenAI ships GPT-5.6 under U.S. oversight and rethinks IPO timing, as a Sikka-led startup bets AI can automate traditional IT services.

Big Tech

OpenAI debuts GPT-5.6 suite under government-reviewed preview

OpenAI is rolling out a new generation of its AI models — GPT-5.6 — in a limited preview that the U.S. government is reviewing customer by customer, The Verge reports. The suite includes three models: Sol (flagship), Terra (mid-tier for high-volume work), and Luna (fast and affordable for everyday use), positioned for coding, cybersecurity, biology, and long-horizon agent-style tasks. ¹

Pricing is per million tokens: GPT-5.6 Sol costs $5 for input and $30 for output, while Terra is roughly half that and Luna is less than half of Terra. Sol also adds two modes — “max” for deeper reasoning and “ultra” for leveraging sub-agents — enabling more complex, multi-step automation. ¹

OpenAI’s announcement centers on safety: The Verge reports the company trained GPT-5.6 to refuse prohibited cyber assistance even under jailbreak attempts, and says Sol is better at finding and fixing vulnerabilities than executing end-to-end attacks under its updated preparedness framework. OpenAI dedicated about 700,000 A100e GPU hours to automated red-teaming and is continuing third-party testing for two weeks. ¹

The preview is tighter than usual, with safeguards that may occasionally interrupt legitimate dual-use work and with case-by-case access approval monitored by the Trump administration, per The Verge. OpenAI says it cooperated with the U.S. government for this launch and aims for broader availability in the coming weeks, while saying such access processes should not be the long-term default. ¹

OpenAI considers delaying IPO to next year

OpenAI is weighing a plan to hold off its initial public offering until next year, the New York Times reported, according to Reuters. Reuters adds the company has confidentially filed for a U.S. IPO targeting a valuation of up to $1 trillion; CFO Sarah Friar has told some associates the company is aiming for a 2027 listing, and advisers outlined options that trade speed for valuation, with CEO Sam Altman resisting cuts to the trillion-dollar target. ²

Reuters also reports the Trump administration asked OpenAI to stagger the release of its new model over security concerns, with Altman telling staff the preview would be approved customer by customer; The Information first detailed that plan. The request came from the Office of the National Cyber Director and the Office of Science and Technology Policy, tying capital-markets timing to heightened policy oversight. ²

Industry & Biz

Vishal Sikka’s Hang Ten raises $32M to rethink services

Hang Ten Systems, a new company led by former Infosys CEO Vishal Sikka, raises a $32 million seed round led by Mayfield with Aramco Ventures participating, TechCrunch reports. The startup says it helps enterprises continuously build, modify, and operate software with AI-driven development and automation, counts Siemens Gamesa Renewable Energy and Fresenius among early customers, and includes Yahoo co-founder Jerry Yang on its board. ³

The launch lands amid a debate over whether AI will disrupt or expand IT services economics: Jefferies argued the sector may see meaningful AI disruption, while Infosys chairman Nandan Nilekani said AI could enlarge the market; Infosys recently framed “AI-first services” as a $300 billion–$400 billion 2030 opportunity. As of Jun 24, 2026, TechCrunch reports investors are reassessing the category, with Infosys shares down over 35% this year. ³

What This Means for You

For teams using AI for coding or security, the GPT-5.6 preview’s tighter safeguards and case-by-case approvals mean some legitimate dual-use work (like vulnerability discovery) may be interrupted. Plan fallbacks, document defensive purpose in prompts, and be ready to route sensitive tasks through approved channels during the preview period. ¹

The new tiering creates room to match model cost to task: run heavy reasoning or high-risk steps on Sol (and “max” when needed), push high-volume routine work to Terra or Luna, and watch your input/output token split since Sol is $5 in / $30 out per million tokens. A simple spreadsheet with your last month’s token mix can reveal quick savings without sacrificing outcomes. ¹

IPO timing talk signals leadership focus on valuation and regulatory coordination, not a shift in product availability today. If you’re negotiating contracts, prefer clear SLAs and termination options over long lock-ins, given the moving parts around oversight and rollout cadence. ²

If you run IT or product operations, Hang Ten’s pitch is a reminder to re-scope repetitive integration and maintenance as AI-automatable micro-projects. Start by tagging a narrow, recurring ticket class (e.g., data mapping or test-case updates) and measure cycle time with an AI assistant versus your current process. ³

Action Items

1. Estimate GPT-5.6 costs by tier: Build a one-page cost model using $5/$30 for Sol and approximate Terra/Luna rates; map at least three recurring tasks to a tier.
2. Request preview access information: Email your OpenAI account team to ask about GPT-5.6 preview eligibility and what evidence helps approvals (e.g., defensive cybersecurity use).
3. Create a “defensive security” prompt template: Draft standard language that states your intent is vulnerability finding/fixing and prepare screenshots/logs to reduce false blocks during review.
4. Pilot an AI-assisted maintenance task: Pick one routine integration, migration, or QA task and run it with your existing AI assistant; record time saved and quality outcomes.
5. Brief your leadership on IPO context: Summarize the reported timing options and confirm contract terms (SLAs, termination) align with your risk tolerance.