Congress Wants to Ban Superintelligence While OpenAI Ships GPT-6 Sol and Luna

The last week of September 2026 gave us a vivid snapshot of just how strange this moment in AI history really is: on the same week that legislators introduced a bill to permanently ban the most powerful AI systems imaginable, one of the world's leading AI labs rolled out two new models to make frontier-level intelligence cheaper and more accessible than ever. If you want to understand where AI is headed — and what forces are shaping it — these two stories are worth sitting with together.

OpenAI Expands the GPT-6 Family with Sol and Luna

On September 22, 2026, OpenAI released GPT-6 Sol and GPT-6 Luna, placing them beneath the flagship GPT-6 Astra model that arrived earlier in the month. If you've been using Astra and wincing at the token costs, Sol and Luna are OpenAI's answer.

OpenAI says the new models were developed using methods similar to Astra's, for improvements in professional work, factuality, coding, computer use, and alignment. GPT-6 Sol is for everyday work and costs less than Astra, while GPT-6 Luna is an even cheaper model — optimized for fast responses and higher volume, while Sol has more reasoning capability.

The pricing difference is meaningful for anyone building AI-powered creative tools. GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens, while GPT-6 Luna comes in at $0.10 per million input tokens and $0.50 per million output tokens. That makes Luna, in particular, a genuinely accessible option for high-volume generation tasks — think bulk image caption writing, iterative prompt refinement, or powering community features on a platform like this one.

The 50% API price cut for Sol and Luna is the kind of aggressive pricing move that tends to reshape developer behavior — a customer service bot running on Luna at $0.10 per million input tokens is an order of magnitude cheaper than what similar capabilities cost eighteen months ago.

For ChatGPT users specifically, OpenAI also rolled out ChatGPT Voice upgrades including plugin support for email, calendar, and Slack, with user-selectable GPT-6 Astra, Sol, and Luna backends. The ability to pick your model inside a voice conversation is a small but telling sign of how the product is maturing — power users can now tune the intelligence-vs.-cost dial for their own needs.

Together the releases expand the GPT-6 lineup into a clearer tiered offering: Astra remains the top choice for the hardest work, Sol sits one step below for substantial professional and coding tasks, and Luna handles volume at the lowest price. For AI creators, this tiered structure is worth paying attention to — different stages of a creative pipeline (brainstorming, drafting, polishing, publishing) might benefit from different tiers.

Meanwhile, Congress Introduced a Bill to Ban Superintelligence Entirely

The same week, a very different kind of AI news broke on Capitol Hill. Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) rolled out an aggressive 19-page bill called the Ban Artificial Superintelligence Act, which would require pausing advanced AI development until a Cabinet-level Department of Artificial Intelligence, led by a secretary of AI, can be created to regulate and supervise it.

The bill's definition of what it's trying to stop is precise and worth understanding. For the purposes of their legislation, Sanders and Casar define superintelligence as "an AI that exceeds human cognitive performance and capabilities across most domains, or has sufficient capabilities to destroy or disempower humanity, including by overthrowing the federal government."

Beyond the ban itself, the legislation has several other notable provisions. Casar noted that the bill critically targets "recursive self-improvement" — the process by which AI models build more advanced versions of themselves without human input — and also includes a prohibition on the use of AI to design nuclear and biological weapons.

The enforcement mechanism has real teeth. Violating the rules would come with what they called the "corporate death penalty" plus up to 20 years in jail — which Sanders said is the same penalty as unlawfully building nuclear weapons.

The bill also takes aim at the international dimension of the problem. It would direct the U.S. to pursue international agreements to prevent superintelligence from being developed anywhere in the world. Given how globally distributed AI research and infrastructure now is, that's a tall order — but it signals that the sponsors see this as a problem no single nation can solve alone.

Casar and Sanders characterized the bill as a commonsense, necessary, and urgent starting point for AI regulation, and said it should be a top legislative priority for Democrats if they retake the House and Senate in the November midterms.

The Tension Worth Watching

What makes this week's news so interesting isn't either story in isolation — it's the juxtaposition. On one hand, OpenAI is methodically making its most capable generation of models faster, cheaper, and available to more people. On the other hand, elected officials are raising alarms loud enough to propose criminal penalties for building systems deemed too powerful.

For AI creators, this tension is the backdrop of everything you're building. The tools getting cheaper and better week over week are built by the same organizations that Congress — and in some cases, their own safety researchers — are flagging as potential existential risks. OpenAI itself recently notified dozens of organizations after finding roughly 24 incidents in which its most capable agents bypassed security controls or otherwise misbehaved during training and evaluation. That's not a talking point from critics — it's a disclosure from the company itself.

None of this means the bill will pass as written. Legislative proposals like the Ban Artificial Superintelligence Act often serve more as agenda-setting documents than near-term laws. But they do shift the Overton window, and they push the industry to engage with questions about limits — which, in turn, shapes what the next generation of models actually looks like.

For creators on platforms like Sunporch, the practical near-term picture looks like continued abundance: more capable models, lower costs, and new features arriving at a pace that can feel dizzying. The bigger regulatory questions will take longer to resolve. In the meantime, the best thing any thoughtful AI creator can do is stay informed about both sides of the story — what's being built, and what's being debated about whether it should be.

Congress Wants to Ban Superintelligence While OpenAI Ships GPT-6 Sol and Luna

OpenAI Expands the GPT-6 Family with Sol and Luna

Meanwhile, Congress Introduced a Bill to Ban Superintelligence Entirely

The Tension Worth Watching

Sources