A coordinated call for slowing artificial intelligence development—triggered by the resignation of Anthropic researcher Jacob Coxon and followed by a public statement from CEO Dario Amodei—has wiped billions from technology stocks, sending the Nasdaq down 0.6% and triggering a 5.9% collapse in semiconductor shares as investors reassess the future of AI infrastructure spending.
Why This Matters
• Semiconductor stocks led losses: Nvidia fell 3.4%, with Intel and AMD both dropping over 4% as markets reacted to the possibility that AI development may face regulatory constraints.
• Geopolitical stakes are high: President Trump rejected any slowdown, framing AI as a "zero-sum competition" with China that the US must win.
• Portuguese investors exposed: PSI-20 listed companies with technology holdings face indirect volatility, while local fund managers reassess tech sector weightings.
• Safety measures are being implemented now: Anthropic and OpenAI are actively rolling out new safeguards—not as completed annual processes, but as current, evolving protocols.
The sequence of events began last week when Jacob Coxon, a researcher at Anthropic who previously worked at OpenAI, resigned, citing escalating safety concerns over rapidly advancing AI systems. He warned that models were approaching thresholds of recursive self-improvement that could lead to uncontrollable risks. Within days, Anthropic CEO Dario Amodei published an essay titled "We Must Pace the Frontier," echoing Coxon’s concerns and calling for deliberate slowing of model capabilities to allow safety mechanisms to catch up. OpenAI’s Sam Altman and Elon Musk of SpaceX soon publicly endorsed the position.
Market Reaction Spreads Across Three Continents
The selling started in Asia before dawn. Tokyo’s Nikkei fell 0.86% to 63,492.99, its lowest level in six weeks, with AI-linked stocks leading declines. SoftBank Group dropped 10.7% after Altman confirmed OpenAI would not pursue an initial public offering this year. South Korea’s Kospi lost 3.26%.
Europe’s Stoxx 600 technology index fell 2.17%, with Dutch chip equipment maker ASML and Germany’s Infineon both posting significant losses. By the time US markets opened, semiconductor futures were already indicating sharp declines.
The PHLX Semiconductor Index (SOX) ended the session down 5.9%. Micron Technology dropped 5.3%, Broadcom fell 4.8%, and Intel lost 5.6%. The sell-off reflected investor anxiety that calls for a development pause—whatever their scientific merit—could translate into reduced capital expenditure on the data centers and hardware that power AI training.
Yet analysts at Deutsche Bank suggested the reaction may be overstated. "It is difficult to imagine that companies would voluntarily pull back while rivals continue to advance," they wrote in a research note, adding that the debate might simply shift the composition of AI investment toward security and governance rather than reduce its overall scale.
What Safety Measures Are Currently Being Implemented
The executives' warnings are not abstract. Both Anthropic and OpenAI have adopted concrete safety measures in early to mid-2026, as part of ongoing updates—not completed annual processes.
Anthropic’s Responsible Scaling Policy, updated in February, prevents the company from training or deploying models capable of catastrophic harm unless specific safeguards are in place. The framework includes ASL-3 security standards that mandate real-time monitoring of model inputs and outputs, strict access controls, and protections against "jailbreaks"—attempts to bypass safety guardrails. These measures are actively being applied across current model development.
OpenAI has similarly restructured its development processes. The company’s Preparedness Framework now tracks advanced AI capabilities that could introduce severe risks, requiring safeguards before development proceeds. Greg Brockman, the company’s president, confirmed in September that OpenAI has deliberately delayed some frontier AI work to strengthen security practices, using their own advanced models to identify and fix vulnerabilities. These are not historical events—they are active, evolving protocols.
Both companies conduct third-party evaluations with organizations like the UK AI Security Institute and have urged US Congress to establish mandatory national security requirements for advanced AI systems, arguing that voluntary commitments are insufficient.
What This Means for Residents
Portuguese investors with exposure to international technology funds or individual US semiconductor positions will see the impact reflected in portfolio valuations. The sell-off affects not just direct holdings but also the technology components within many pension funds and indexed savings products popular in Portugal.
More broadly, any slowdown in AI infrastructure investment would affect the demand pipeline for data center components—business lines that European companies like ASML have relied on for growth. For those tracking the sector, the key question is whether this moment represents a genuine inflection point or a temporary correction in a multi-year investment cycle.
Portugal’s regulatory environment already aligns with European Union AI governance frameworks that prioritize risk assessment and accountability. The EU AI Act, enforced as a compliance requirement in 2026, mandates transparency and safety protocols that parallel what US companies are now advocating. Portuguese businesses developing or deploying AI systems should note that the conversation in Washington has moved from theoretical risk to immediate, real-world implementation—a direction European regulation has already taken.
The Geopolitical Standoff That Won’t Allow a Pause
The practical obstacle to any coordinated slowdown lies in Washington and Beijing.
President Trump dismissed the executives’ concerns during remarks in Doonbeg, Ireland, labeling them "negative forces" and rejecting any suggestion that AI development should be constrained. "Whoever wins AI, wins," he stated, framing the technology as a competitive battleground where the United States cannot afford to cede ground to China.
Beijing’s response was equally dismissive. Foreign Ministry spokesman Guo Jiakun rejected the slowdown calls as "alarmism" and "malicious competition," emphasizing that all parties should work together for open and inclusive AI development. China views the proposals with suspicion, suspecting they may be designed to constrain Chinese technological advancement.
Both governments are scheduled to hold talks on AI security in mid-September, with Treasury Secretary Scott Bessent leading the US delegation ahead of a potential Trump-Xi meeting later in the month. But Bessent’s own assessment was blunt: "We cannot stop. We cannot, because the Chinese will not stop."
The dissonance is stark. At precisely the moment when the industry’s most powerful figures are calling for restraint, the world’s two largest powers are intensifying their competition. Markets are left pricing a future where safety concerns may reshape investment priorities, but where the underlying race for technological supremacy shows no sign of slowing.
For investors and observers in Portugal, the coming weeks will reveal whether this moment marks the beginning of a more regulated AI era or simply a speed bump in an accelerating race. The answer will determine whether Tuesday’s semiconductor sell-off was a warning or an opportunity.
