California Pirate Party | May 2026
There’s a version of accountability where companies that cause harm face consequences. And then there’s California’s current approach to AI regulation, where companies that choose to meet safety standards get rewarded with legal protections, and companies that don’t — face nothing.
Welcome to SB 813, the Voluntary AI Standards Act, which passed the California Senate 31-7 in January and is now moving through the Assembly. Its supporters call it a “novel approach.” We have a different word for it: industry capture with better branding.
How We Got Here
To understand how California ended up with voluntary AI regulation, you have to understand what we almost had instead.
In 2024, Senator Scott Wiener introduced SB 1047 — a serious bill. It would have required safety testing and annual third-party audits for the most powerful AI systems. It would have mandated a “kill switch.” It would have created CalCompute, a public computing infrastructure to democratize AI access for researchers, startups, and community groups rather than leaving it entirely to the cloud giants.
The tech industry responded by spending millions to kill it. Meta dropped $4.13 million lobbying California lawmakers. Google spent $2.39 million. The California Chamber of Commerce — which dutifully carries water for every entrenched industry that funds it — spent $11.48 million in lobbying from January through September 2024 alone.
Governor Newsom vetoed SB 1047 in September 2024, calling it not “informed by an empirical trajectory analysis.” He did not define what that analysis would look like, who would conduct it, or when it might be complete. The message was clear enough: Silicon Valley asked, Sacramento delivered.
What We Got Instead
In place of SB 1047’s teeth, California passed:
- SB 53 — Frontier AI developers must test and plan for catastrophic risks. “Catastrophic” is defined as 50 or more deaths, or $1 billion in damages. A bar so high it applies to almost nothing currently deployed.
- AB 853 — AI-generated content must be labeled. Good. Enforceable only after the fact, and only if someone catches it.
These are disclosure laws. They tell you a decision was made by AI. They don’t tell you how to challenge it, who is liable when it’s wrong, or what happens when a company ignores the requirement.
Disclosure without accountability is a press release, not a regulation.
SB 813: The Fig Leaf Matures
Now comes SB 813, dressed up in the language of rigor — commissions, independent verification organizations, expert panels, safety standards. It sounds like regulation.
Here’s what it actually does: companies that voluntarily choose to meet the commission’s standards get certified. Certified companies receive legal protections — a liability shield. Companies that don’t bother with certification receive… nothing. No penalty. No bar to doing business. No consequence whatsoever.
Read that again. The incentive structure of SB 813 is:
- Meet voluntary standards → get protected from lawsuits
- Ignore voluntary standards → continue operating exactly as before
This is not safety regulation. This is a fast lane for companies that can afford to play ball, while leaving the actual problem — unaccountable AI systems making high-stakes decisions — completely unaddressed. It doesn’t matter whether the commission’s standards are rigorous if the only consequence of ignoring them is… not getting a certificate.
Critics have already noted that the bill “falls short in providing the clarity and consistency developers and policymakers need” and risks “inadvertently giving coercive power to organizations that are meant to be voluntary.” Even on its own terms, it doesn’t work. But its own terms are the real problem.
Who Pays When the Opt-Outs Win?
AI systems are already deciding parole recommendations, insurance eligibility, hiring decisions, loan applications, and child welfare interventions in California. The people most affected by those decisions — disproportionately lower-income Californians, people of color, people without lawyers — have no meaningful recourse when the system is wrong.
SB 813 does nothing for them. A certification badge on a company’s website does not give a wrongly denied applicant a path to appeal. It does not tell a community why an algorithm flagged them. It does not hold a developer liable when a model causes harm.
The companies lobbying for voluntary frameworks know exactly what they’re doing. Voluntary standards are the oldest move in the regulatory capture playbook: propose a process rigorous enough to generate good press, structured so participation is optional, and make sure non-participation carries no cost. Then point to the process as evidence that the industry is self-governing. Repeat until the political moment passes.
What Real Accountability Looks Like
The Pirate Party’s position is simple: if you deploy a system that makes consequential decisions about people’s lives, you are responsible for what that system does. Accountability is not optional. Audits should be mandatory. Liability should exist. Affected people should have the right to explanation and appeal — not because a company chose to volunteer for a certification program, but because they are owed it.
SB 1047 wasn’t perfect. But it was a genuine attempt at a mandatory framework with real enforcement mechanisms. CalCompute was a genuine attempt to prevent AI infrastructure from becoming a purely private oligopoly.
Both were killed by lobbying money. What survived is the appearance of regulation — enough to claim the legislature acted, not enough to inconvenience the companies that fund the campaigns.
California can lead on AI policy. What it’s doing right now is following the industry’s preferred script. The difference matters — especially to the people who will be on the receiving end of the systems that never got audited.
The California Pirate Party supports mandatory, enforceable AI accountability; the right to explanation and appeal; and public AI infrastructure that isn’t controlled by the same companies it’s supposed to regulate.
What You Can Do
SB 813 is still in the Assembly. It hasn’t passed yet.
1. Contact your Assembly member. Tell them voluntary AI standards are not AI safety. Ask them to amend SB 813 to include mandatory compliance for high-stakes AI deployments, or vote no. Find your rep at findyourrep.legislature.ca.gov.
2. Ask Newsom where the line is. He vetoed SB 1047 as too broad. SB 813 is so narrow it’s meaningless. His office takes public comment: gov.ca.gov/contact. Ask him what mandatory AI accountability would actually need to look like for him to sign it.
3. Support organizations doing the work. The Electronic Frontier Foundation, Algorithmic Justice League, and the Transparency Coalition are tracking these bills and pushing for real enforcement. They need visibility and members more than money.
4. Show up to hearings. Assembly committee hearings on SB 813 are public. Lobbyists fill those rooms. Constituents mostly don’t. That asymmetry is part of how we got here. leginfo.legislature.ca.gov has the hearing schedule.
5. Talk about it. Most Californians don’t know that AI is already making decisions about their parole, insurance, and job applications — with no mandatory audit requirement and no right of appeal. That information gap is itself a form of power. Close it.
The companies that killed SB 1047 are counting on this staying a wonky legislative story that nobody outside Sacramento pays attention to. Prove them wrong.
California Pirate Party | May 2026
There’s a version of accountability where companies that cause harm face consequences. And then there’s California’s current approach to AI regulation, where companies that choose to meet safety standards get rewarded with legal protections, and companies that don’t — face nothing.
Welcome to SB 813, the Voluntary AI Standards Act, which passed the California Senate 31-7 in January and is now moving through the Assembly. Its supporters call it a “novel approach.” We have a different word for it: industry capture with better branding.
How We Got Here
To understand how California ended up with voluntary AI regulation, you have to understand what we almost had instead.
In 2024, Senator Scott Wiener introduced SB 1047 — a serious bill. It would have required safety testing and annual third-party audits for the most powerful AI systems. It would have mandated a “kill switch.” It would have created CalCompute, a public computing infrastructure to democratize AI access for researchers, startups, and community groups rather than leaving it entirely to the cloud giants.
The tech industry responded by spending millions to kill it. Meta dropped $4.13 million lobbying California lawmakers. Google spent $2.39 million. The California Chamber of Commerce — which dutifully carries water for every entrenched industry that funds it — spent $11.48 million in lobbying from January through September 2024 alone.
Governor Newsom vetoed SB 1047 in September 2024, calling it not “informed by an empirical trajectory analysis.” He did not define what that analysis would look like, who would conduct it, or when it might be complete. The message was clear enough: Silicon Valley asked, Sacramento delivered.
What We Got Instead
In place of SB 1047’s teeth, California passed:
These are disclosure laws. They tell you a decision was made by AI. They don’t tell you how to challenge it, who is liable when it’s wrong, or what happens when a company ignores the requirement.
Disclosure without accountability is a press release, not a regulation.
SB 813: The Fig Leaf Matures
Now comes SB 813, dressed up in the language of rigor — commissions, independent verification organizations, expert panels, safety standards. It sounds like regulation.
Here’s what it actually does: companies that voluntarily choose to meet the commission’s standards get certified. Certified companies receive legal protections — a liability shield. Companies that don’t bother with certification receive… nothing. No penalty. No bar to doing business. No consequence whatsoever.
Read that again. The incentive structure of SB 813 is:
This is not safety regulation. This is a fast lane for companies that can afford to play ball, while leaving the actual problem — unaccountable AI systems making high-stakes decisions — completely unaddressed. It doesn’t matter whether the commission’s standards are rigorous if the only consequence of ignoring them is… not getting a certificate.
Critics have already noted that the bill “falls short in providing the clarity and consistency developers and policymakers need” and risks “inadvertently giving coercive power to organizations that are meant to be voluntary.” Even on its own terms, it doesn’t work. But its own terms are the real problem.
Who Pays When the Opt-Outs Win?
AI systems are already deciding parole recommendations, insurance eligibility, hiring decisions, loan applications, and child welfare interventions in California. The people most affected by those decisions — disproportionately lower-income Californians, people of color, people without lawyers — have no meaningful recourse when the system is wrong.
SB 813 does nothing for them. A certification badge on a company’s website does not give a wrongly denied applicant a path to appeal. It does not tell a community why an algorithm flagged them. It does not hold a developer liable when a model causes harm.
The companies lobbying for voluntary frameworks know exactly what they’re doing. Voluntary standards are the oldest move in the regulatory capture playbook: propose a process rigorous enough to generate good press, structured so participation is optional, and make sure non-participation carries no cost. Then point to the process as evidence that the industry is self-governing. Repeat until the political moment passes.
What Real Accountability Looks Like
The Pirate Party’s position is simple: if you deploy a system that makes consequential decisions about people’s lives, you are responsible for what that system does. Accountability is not optional. Audits should be mandatory. Liability should exist. Affected people should have the right to explanation and appeal — not because a company chose to volunteer for a certification program, but because they are owed it.
SB 1047 wasn’t perfect. But it was a genuine attempt at a mandatory framework with real enforcement mechanisms. CalCompute was a genuine attempt to prevent AI infrastructure from becoming a purely private oligopoly.
Both were killed by lobbying money. What survived is the appearance of regulation — enough to claim the legislature acted, not enough to inconvenience the companies that fund the campaigns.
California can lead on AI policy. What it’s doing right now is following the industry’s preferred script. The difference matters — especially to the people who will be on the receiving end of the systems that never got audited.
The California Pirate Party supports mandatory, enforceable AI accountability; the right to explanation and appeal; and public AI infrastructure that isn’t controlled by the same companies it’s supposed to regulate.
What You Can Do
SB 813 is still in the Assembly. It hasn’t passed yet.
1. Contact your Assembly member. Tell them voluntary AI standards are not AI safety. Ask them to amend SB 813 to include mandatory compliance for high-stakes AI deployments, or vote no. Find your rep at findyourrep.legislature.ca.gov.
2. Ask Newsom where the line is. He vetoed SB 1047 as too broad. SB 813 is so narrow it’s meaningless. His office takes public comment: gov.ca.gov/contact. Ask him what mandatory AI accountability would actually need to look like for him to sign it.
3. Support organizations doing the work. The Electronic Frontier Foundation, Algorithmic Justice League, and the Transparency Coalition are tracking these bills and pushing for real enforcement. They need visibility and members more than money.
4. Show up to hearings. Assembly committee hearings on SB 813 are public. Lobbyists fill those rooms. Constituents mostly don’t. That asymmetry is part of how we got here. leginfo.legislature.ca.gov has the hearing schedule.
5. Talk about it. Most Californians don’t know that AI is already making decisions about their parole, insurance, and job applications — with no mandatory audit requirement and no right of appeal. That information gap is itself a form of power. Close it.
The companies that killed SB 1047 are counting on this staying a wonky legislative story that nobody outside Sacramento pays attention to. Prove them wrong.