Altman presses Washington on AI shocks

Sam Altman has urged Washington to prepare for the upheaval and promise of what he describes as a coming era of AI “superintelligence”, arguing that the United States needs a broader policy response to manage job disruption, cyber threats and the concentration of wealth that advanced systems could create.

The OpenAI chief executive’s latest intervention comes in a new policy blueprint that says incremental regulation will not be enough as AI systems grow more capable. The document argues that stronger public institutions, wider access to computing power and a new social compact will be needed if the gains from AI are not to flow mainly to a small cluster of companies and investors.

Altman’s position reflects a widening shift in Silicon Valley, where executives who once focused on model performance and commercial adoption are now speaking more openly about the economic and political consequences of highly capable AI. OpenAI’s paper says the country should think beyond narrow rules for individual tools and instead prepare for structural change in labour markets, taxation, public services and national resilience. Among the ideas advanced are a public wealth fund, stronger safety nets, a tax system less dependent on labour income and policies to spread the benefits of productivity gains more broadly across society.

That message arrives at a moment when AI policy is becoming entangled with industrial strategy. OpenAI has already been pushing governments to support more data centres, energy capacity and wider deployment of AI in public services and business. The latest blueprint adds a sharper warning that the technology’s upside could be matched by destabilising side effects if governments fail to move early.

One of the most immediate pressure points is cybersecurity. As AI tools make it easier to generate code, scan for weaknesses and automate parts of the attack cycle, security specialists are warning that the barrier to carrying out sophisticated digital attacks is falling. That matters acutely in crypto markets, where decentralised systems can expose large pools of capital to operational mistakes, private-key theft and software flaws. A growing concern across the sector is that the same AI tools that help developers work faster can also accelerate the discovery and exploitation of vulnerabilities.

The scale of the threat is already visible in the numbers. Chainalysis said roughly $2.2 billion was stolen from crypto platforms in 2024, up about 21 per cent from the year before, while TRM Labs said hackers had already taken about $1.38 billion by June 24 of that year, driven by a small number of large incidents. Chainalysis also found that private-key compromises accounted for the largest share of stolen funds in 2024, underscoring how attacks are increasingly targeting operational infrastructure rather than only smart-contract bugs.

By 2025, the pattern had become even more severe. TRM Labs said illicit actors stole $2.87 billion across nearly 150 hacks and exploits in 2025, with attackers shifting “up the stack” towards wallets, keys and control systems. That trend is central to the debate over AI risk because tools that lower the expertise needed to probe code or imitate developer workflows may expand the number of capable attackers, even if the largest breaches still depend on human planning and access.

Altman’s remarks also land amid an increasingly public argument over the speed of AI development. OpenAI has been signalling for months that it sees the path to more powerful systems as advancing quickly, while critics and rivals have warned that governance, safety testing and public oversight are not keeping pace. The company’s own language has become more direct, saying society is moving towards superintelligence and that the transition may require something closer to a national economic strategy than a traditional tech rulebook.

For Washington, the political challenge is that AI is now being framed in two competing ways at once: as a growth engine that could strengthen productivity, science and national power, and as a source of labour displacement, market concentration and cyber vulnerability. Altman is trying to place himself on both sides of that argument, warning of disruption while also urging the United States to build the infrastructure and institutions needed to lead.

Whether lawmakers embrace that message may depend on how quickly abstract concerns turn into visible economic strain. The more AI is woven into coding, finance, logistics and public administration, the harder it will be to treat the technology as just another fast-growing industry. What is becoming clearer is that the debate has moved beyond chatbots and model launches. It is now centred on who controls the gains, who bears the risk and whether democratic institutions can keep pace with machines that their creators say may soon surpass human capability in many domains.