‘Insane’ demand and security jitters collide as OpenAI’s Sol races from DC review into the office

• Government scrutiny, then a green light
• A workplace ‘super app’ and the paradox of choice
• ‘Insane’ growth, mounting strain
• Enthusiasm, competition, and unresolved fears

‘Insane’ demand and security jitters collide as OpenAI’s Sol races from DC review into the office
OpenAI’s most powerful new AI model, Sol, is rocketing into offices worldwide just weeks after a federal review over its cybersecurity risks, creating a clash between soaring demand and lingering unease about what it can actually do.

§Government scrutiny, then a green light

In June, OpenAI quietly debuted the GPT-5.6 family — Sol, Terra, and Luna — but kept public access on hold while the Trump administration reviewed the models’ advanced cybersecurity capabilities, similar to the scrutiny faced by Anthropic’s Fable 5. GPT-5.6 is described as OpenAI’s “strongest cybersecurity model yet,” supporting threat modeling, code review, patching, and internal “blue team” attack simulations that can probe systems for weaknesses before real hackers do.

On July 9, after the government’s go-ahead, OpenAI officially rolled out GPT-5.6 and unveiled ChatGPT Work, an AI agent that mixes ChatGPT and Codex-style automation for non-coders and professionals.

§A workplace ‘super app’ and the paradox of choice

OpenAI is folding its Codex coding tool into a unified ChatGPT desktop app, pitched as a “one-stop shop” for engineers and office workers. ChatGPT Work can “stay with a project for hours if needed, and turn a goal into finished work,” from budget analysis to marketing campaigns, while integrating with tools like Slack, Teams, and Google Drive.

But the power comes with complexity. Users now must pick among Sol, Terra, and Luna, and tune “reasoning” levels to balance speed, quality, and cost — a “Goldilocks” problem that makes maximizing productivity non-trivial for non-experts.

§‘Insane’ growth, mounting strain

As the rollout accelerated globally, partners like Amazon announced that GPT-5.6 is “now generally available on Bedrock,” offering three tiers of intelligence on its high-performance infrastructure. Business coverage framed the move as OpenAI’s biggest push yet to make ChatGPT indispensable at work, with nearly 1 billion users in its ecosystem.

Demand quickly spiked. Axios reported that CEO Sam Altman warned Sol’s “growth is insane” and that, despite “heroic work” from the inference team, “it is possible there are some hiccups soon.” The comment underscored how even a top AI vendor can struggle to scale compute as rivals like Anthropic, Meta, and SpaceXAI launch competing frontiers.

§Enthusiasm, competition, and unresolved fears

Benchmarks and early testers suggest Sol may now be the leading general AI model, with OpenAI citing frontier performance in cybersecurity and coding efficiency, and outside evaluators placing Sol at the top of design and reasoning leaderboards. Yet the same capabilities that excite developers and enterprise vendors prompted federal intervention in the first place, and critics note that handing ChatGPT Work broad access to corporate files and apps raises its own security and governance risks.

As Sol spreads through cloud platforms and productivity suites, the core tension remains: AI agents are becoming more capable, persistent, and embedded in everyday workflows, even as policymakers and users are still working out how — and whether — they can be safely trusted.

Continue reading https://foxvector.com/stories/019f675c-e405-3326-707b-0d87460950b7