Opinion by Josh Hamilton. Reviewed 2 October 2026. The original publication date is preserved; the GPT-5.6 availability claim has been corrected. The access argument below is a policy concern, not a prediction that exclusion is inevitable.
For the last few years, artificial intelligence has felt like one of the few major technological shifts that actually had the potential to level the playing field. A small business owner, a regional operator, a sole trader, or a local service provider could suddenly access tools that were once only available to large corporations with internal tech teams, consultants, and enormous budgets. That mattered, because AI was not just making big companies more powerful. It was also giving smaller businesses a chance to compete, improve their systems, and close the gap. But recent events should make all of us pause. On 12 June 2026, Anthropic said a U.S. government directive targeting access by foreign nationals required it to suspend Fable 5 and Mythos 5 for all customers. Other Anthropic models were unaffected, according to the company. That is Anthropic’s account of the directive, rather than an independently published government explanation. Read Anthropic’s statement.
OpenAI took a different path: after a limited preview, it announced general availability of GPT-5.6 on 9 July 2026 across ChatGPT, Codex and its API. By this article’s original publication date, it was therefore inaccurate to describe GPT-5.6 as still confined to a small preview group. OpenAI did retain additional access controls for sensitive cyber capabilities. Read OpenAI’s launch announcement.
These are different decisions, and they do not establish that frontier AI will permanently become an exclusive club. They do raise a question worth asking: how should security restrictions preserve a fair path to access? And to be clear, I do not think security concerns around frontier AI should be dismissed. Powerful AI systems can create real risks, particularly around cyber capability, automation, and misuse. The problem is not that governments and AI companies are thinking about safety. They absolutely should be. The problem is what happens if safety becomes the justification for a future where only selected organisations, major corporations, government-approved partners, and strategically important institutions get access to the most capable models first. That is where this starts to become uncomfortable. Because once access to advanced AI becomes selective, the existing power imbalance gets worse. Big companies already have the advantage. They have bigger budgets, more staff, better infrastructure, stronger legal teams, deeper vendor relationships, and more influence. If they are also the first to receive the most capable AI systems, they get another head start while everyone else is told to wait for the public version, the safer version, the restricted version, or the version deemed suitable for ordinary users. That might sound reasonable from a national security perspective, but from a business perspective it creates a serious access divide. My concern is whether small and medium businesses have enough of a voice when these decisions are made. A local mechanic, electrical contractor, real estate office, medical clinic, transport company, accounting firm, retailer, or regional manufacturer needs a practical route to capable tools, even without a direct relationship with a frontier AI lab. Delayed access could affect their competitiveness. Their interests deserve explicit consideration. And that is exactly the issue. AI is not just a novelty anymore. It is becoming business infrastructure. It is being used for administration, customer service, quoting, scheduling, reporting, workflow automation, marketing, training, compliance, data analysis, and decision support. For a small business, access to better AI does not mean playing with shiny tech. It can mean fewer missed calls, faster responses, better follow-up, reduced admin pressure, better use of staff time, and stronger customer service. So when the most capable systems are restricted to a narrow group of trusted users, we need to ask who benefits from that arrangement. Because if the largest companies are allowed to build with the best tools first, while small businesses are left waiting, then we are not democratising technology. We are creating another layer of advantage for the people and organisations who already had the most. There is also a practical flaw in the idea that restricting access automatically solves the security problem. Access controls cannot be the whole security strategy. We also need to consider misuse through stolen credentials, leaked systems or alternative tools. Meanwhile, legitimate businesses, researchers, developers, educators, and cyber defenders may be left operating with weaker access and fewer options. That does not mean every advanced model should be released recklessly to everyone on day one. That would be naive. Some level of staged rollout, monitoring, and safeguards makes sense. But there is a difference between responsible deployment and quiet gatekeeping. Responsible deployment should mean clear rules, transparent criteria, proper safeguards, user education, strong monitoring, and a fair pathway for legitimate users to gain access. Gatekeeping looks very different. It looks like vague approval processes, private lists, trusted insiders, government-approved favourites, and everyone else being told to trust the process. That is the part we should be careful about accepting too easily. Because if this becomes the normal pattern, then the future of AI may not halt completely, but broad access to progress might. The technology will keep improving, but the benefits may flow first to the organisations already closest to power. Everyone else will get access later, once the early advantage has already been captured. For regional businesses, that matters. Many are already at risk of being overlooked in the digital economy. They do not always have the time, money, or internal expertise to chase every new tool or technical shift. AI could help them catch up, modernise, and compete. But only if they can access capable tools in a fair and practical way. At Hamilton Lockhart, my focus is helping small and medium businesses understand and adopt AI in a way that is useful, safe, and realistic. Not hype. Not Silicon Valley theatre. Not pretending every business needs some massive enterprise transformation project. Practical AI is about helping real businesses improve their workflows, reduce repetitive admin, support their teams, and stay competitive in a changing economy. But that mission depends on access. If the most capable AI tools are only made available to a select group of approved organisations, then small businesses will once again be forced to compete from behind. They will be asked to adapt to an AI-driven economy while being given delayed or limited access to the very tools shaping that economy. Security matters. It always will. But access matters too. We should not accept a future where the best AI systems are quietly reserved for the elite while everyone else is handed the leftovers later. If AI is going to reshape business, productivity, and opportunity, then small businesses deserve more than delayed access and second-tier tools. They deserve a seat at the table. And if they are not given one, then the rest of us will have to keep building one.