US Government Tightens OpenAI Access Under Sam Altman as Bank of England Flags AI Agents as Financial Stability Risk

The developments highlight a broader shift in how governments and central banks are approaching advanced AI. Rather than focusing solely on innovation, regulators are increasingly weighing the technology’s implications for national security, financial markets, and critical infrastructure.

US Government reportedly limits OpenAI GPT-5.6 rollout

According to @ns123abc, the Trump administration requested that OpenAI introduce its upcoming GPT-5.6 model through a limited preview instead of a broad public launch. The report states that access would initially be restricted to selected partners, with additional users receiving approval on a case-by-case basis due to security considerations.

The Trump administration reportedly asked OpenAI to roll out GPT-5.6 as a limited preview for select partners, with government approval granted on a customer-by-customer basis over security concerns. Source: @ns123abc via X

OpenAI CEO Sam Altman reportedly indicated that access to the model would be approved “customer by customer,” suggesting a more controlled deployment process than previous flagship releases.

Although neither the White House nor OpenAI has publicly outlined a permanent licensing framework for frontier AI models, the reported approach has sparked debate across the technology industry. Some observers argue that tighter government oversight could help reduce security risks associated with increasingly capable AI systems. Others warn that limiting access may slow domestic innovation while giving overseas competitors additional room to advance.

Ben Cera predicts China’s open-source AI models, including GLM-5.2, could sharply reduce AI service costs, pressuring the valuations of companies like OpenAI and Anthropic. Source: Ben Cara via X

The discussion comes as competition in the global AI race continues to intensify. Chinese developers have recently introduced increasingly capable open-weight models, with systems such as GLM-5.2 attracting attention for achieving strong performance on coding, software engineering, and agentic AI benchmarks at significantly lower operating costs than many frontier proprietary models.

Industry analysts note that while lower-cost models could place pressure on premium AI pricing, enterprise customers still evaluate AI platforms based on broader considerations including reliability, cybersecurity, regulatory compliance, infrastructure, and integration capabilities.

Bank of England warns AI agents could create financial stability risk

Separately, Bank of England Deputy Governor Sarah Breeden cautioned that autonomous AI agents could become a growing source of financial instability as they assume more decision-making responsibilities across financial markets.

Bank of England Deputy Governor Sarah Breeden warned that autonomous AI agents could amplify market volatility and pose financial stability risks through synchronized trading behavior. Source: Bloomberg via X

Speaking at the European Central Bank’s annual forum in Sintra, Portugal, on June 30, Breeden said AI capabilities are advancing faster than many regulators had anticipated.

“We were surprised this Spring, and we should be prepared for further technology surprises,” she said.

Breeden explained that AI development has progressed rapidly over the past several years. Early generative AI systems primarily produced content after receiving user prompts. More recent reasoning models learned to solve complex, multi-step problems, while the latest generation of agentic AI can independently plan and execute sequences of actions with minimal human involvement.

She said these capabilities could eventually allow AI agents to execute securities trades, process payments, and respond to cybersecurity incidents autonomously, creating a financial ecosystem that “operates more autonomously, at scale and speed.”

According to Breeden, synchronized behavior among multiple AI systems presents a particular concern. If numerous agents respond similarly to identical prompts or develop comparable objectives, their actions could amplify market volatility and potentially accelerate financial disruptions.

Cybersecurity and AI investment risks remain key concerns

Breeden identified cybersecurity as one of the most immediate threats associated with increasingly capable AI systems. Citing findings from the UK’s AI Security Institute, she noted that agentic AI has become significantly more effective at identifying software vulnerabilities.

While these capabilities can strengthen cyber defenses by helping organizations detect weaknesses faster, they may also enable malicious actors to discover and exploit vulnerabilities more efficiently.

“The malicious use of these capabilities materially increases the chance of attacks that could harm financial stability,” Breeden warned.

She also suggested that highly autonomous trading systems may eventually require built-in safeguards, including potential “kill switches,” to reduce the risk of cascading market disruptions during periods of extreme volatility.

Beyond operational risks, Breeden highlighted growing financial vulnerabilities tied to the AI investment boom. The Bank of England’s Financial Policy Committee previously observed that major technology firms are increasingly relying on more complex debt financing to fund AI infrastructure after initially depending largely on cash flow and equity.

According to the committee, a sharp decline in AI-related asset values could have broader consequences for credit markets as leverage across the sector continues to increase. A more detailed assessment of these risks is expected in the committee’s upcoming review.

Breeden also argued that central banks should not focus exclusively on regulating AI-related risks. Instead, she said supervisory authorities should begin adopting AI tools themselves to improve oversight and better understand increasingly automated financial systems.

Together, the reported U.S. restrictions on advanced AI deployments and the Bank of England’s warnings underscore a common theme emerging among policymakers: as AI systems become more autonomous and influential, governments are moving toward stronger oversight aimed at balancing technological innovation with economic resilience, cybersecurity, and financial stability.