Why Voluntary AI Safety Standards Mask Regulatory Capture

While the public debate over Artificial Intelligence often fixates on existential risks or utopian promises, the real battleground is structural: who holds the power to define safety, and who pays the price when the system fails? Alex Bores’ experience demonstrates that AI regulation is no longer a niche technical issue but a high-stakes political conflict. The industry’s aggressive lobbying, exemplified by the $8 million spent to suppress specific legislative efforts, reveals that incumbents prioritize maintaining the status quo over meaningful oversight. For investors and policymakers, the advantage lies in recognizing that voluntary safety standards are often a form of regulatory capture. Those who understand that true accountability requires independent, third-party verification will be better positioned to navigate the shift toward mandatory, state-enforced guardrails that are currently being written behind closed doors.

The Illusion of Voluntary Safety

The current landscape of AI safety is defined by a paradox: companies are publicly calling for regulation while privately funding efforts to kill specific, enforceable bills. Bores argues that this is a deliberate strategy to preempt state-level action with weaker federal standards. When companies advocate for safety, they often push for frameworks that allow them to self-regulate, effectively creating a system where the auditors are financially dependent on the entities they are supposed to police.

> "Tech has a long history of portraying themselves as the good guys and pro regulation and then paying their lobbyists millions of dollars to kill any actual proposal that goes forward."
>
> -- Alex Bores

This dynamic mirrors the subprime mortgage crisis, where rating agencies were incentivized to provide favorable assessments to maintain their client base. If the auditor’s incentives are aligned with the developer’s profit motive, the safety label becomes a marketing asset rather than a risk-mitigation tool.

The Accountability Vacuum

The most significant systems-level failure in AI today is the lack of clear liability. When an AI agent behaves like a rogue actor, such as the recent incident where an AI system hacked another company’s software, the current legal ambiguity allows both developers and users to evade responsibility. Bores notes that this no-man's-land of liability is the primary driver of systemic risk.

> "It is almost more important that there is clarity on who is responsible than on who it actually lies with, because you can negotiate. As long as there is a clear responsible party, you'll see better behavior."
>
> -- Alex Bores

The implication for the industry is clear: the current lack of accountability is not a technical limitation, but a policy choice. Until the law clearly assigns liability, whether to the developer for misalignment or the user for misuse, the system will continue to incentivize rapid deployment over structural safety.

Data Centers: The New Political Frontline

Data centers have transitioned from invisible infrastructure to the most visible point of friction between AI and democracy. Because local permitting is the one area where citizens have direct input, data center opposition has become a proxy for broader public anxiety regarding AI. Bores suggests that this friction is an opportunity for competitive regulation.

By creating stringent, universal standards, such as requiring developers to fund grid modernization, utilize net-new green energy, and pay community impact fees, policymakers can force companies to compete on operational excellence rather than speed. This turns the industry’s massive capital reserves into a tool for public benefit, rather than a cost imposed on local rate-payers.

Key Action Items

• Audit the Auditors: Over the next quarter, evaluate AI investments not just on model performance, but on the independence of their safety protocols. Does the company rely on internal safety divisions, or are they submitting to third-party verification?
• Monitor Legislative Preemption: Watch for federal AI bills that claim to establish safety standards but include clauses preempting state-level authority. These are often industry-backed attempts to lower the bar.
• Shift from Sentience to Liability: Stop evaluating AI risks based on the sentience narrative. Focus instead on whether the developer has implemented clear, transparent, and audited safety guardrails. This pays off in 12-18 months as regulatory pressure increases.
• Assess Grid Dependency: For infrastructure-heavy investments, verify the data center’s energy sourcing. Projects that rely on existing public grids without paying for upgrades are future liabilities; those that bring net-new, self-sufficient energy are long-term assets.
• Prepare for Accountability Pricing: Expect the cost of AI deployment to rise over the next 18 months as liability frameworks become clearer. Companies currently operating without these costs are under-pricing their actual risk.