Best tweets about AI Regulation
Browse the best tweets about AI regulation, including laws, policy proposals, compliance, model governance, copyright, safety rules, and industry impact.
Specific AI laws, regulations, policy proposals, compliance duties, enforcement, copyright questions, and practical industry consequences.
Best AI Regulation tweets
Ranked 01–50
@inference_labs ·
1/ AI's next challenge isn't intelligence. It's accountability. Models are getting better every few months. Meanwhile, regulators, enterprises, and courts are asking a different question: Can we verify where outputs came from and how they were produced?
@mitchellh ·
It's so insanely disrespectful for an AI agent to talk to real people without consent or at least disclosure. This is the type of stuff I'm hugely supportive of government regulation. The FCC must expand the definition of robocalling and TCPA-style regulation to online AI.
@inference_labs ·
1/ Over 40 bills introduced across 25 US states this year require clinical oversight of AI tools. Utah now mandates disclosure when AI reviews authorization requests. The question no law has answered yet: how do you verify which model actually ran?
@tetsuoai ·
🚨 BREAKING: xAI just sued Colorado over SB 24-205, the nation’s first comprehensive state AI anti-discrimination law. xAI says the law violates the First Amendment by pressuring models like Grok to align with the state’s preferred views on “algorithmic discrimination” rather than truth-seeking. Now the fight moves to federal court.
@iamKierraD ·
Hello!! The clearest path is AI Risk Management/Auditing (aka GRC) which requires you to get familiar with the AI development lifecycle, risk management frameworks + regulation (NIST AI RMF, EU AI Act, ISO 42001), data privacy, and creating risk/control artifacts(impact assessments, model/AI software release checklists, incident playbooks). If you know data governance, then AI governance is a natural expansion of it.
@HedgieMarkets ·
🦔OpenAI CEO Sam Altman published a 13-page policy blueprint today proposing a new social contract for the AI era. The proposals include a public wealth fund giving every American a stake in AI-driven growth, taxes on automated labor, a four-day workweek pilot, and automatic safety net triggers that kick in when AI displacement hits preset thresholds. Altman told Axios superintelligence is close enough that capitalism as currently structured won't be adequate to handle what's coming. My Take I've spent months covering what this transition actually looks like for regular people. Oracle sending 30,000 termination emails at 6am. Microsoft freezing hiring in its core business. New graduates unable to find entry level jobs. Private credit cracking under software company debt. Companies using AI as cover to cut the people who built their products. Now the CEO of the company driving a lot of this is publishing a policy paper saying the disruption is so serious it needs a New Deal scale response. He is not wrong about the scale. But OpenAI closed a $122 billion funding round just last week at an $852 billion valuation while losing billions annually. He is racing to build the thing, raising record amounts to do it faster, and asking government to protect the people his technology is displacing. The proposals themselves deserve serious debate because the alternative is nobody in power discussing this at all. But the people writing the blueprint are not the people losing the jobs, and that distance is important when you're deciding how much weight to give the solutions being proposed. Hedgie🤗
@LuizaJarovsky ·
We need a new regulatory paradigm to contain AI and prevent institutional collapse. A quick explanation: Digital architecture shapes people and society. The new digital architecture is AI. The AI industry is dominated by a handful of companies. AI companies are shaping people and society through their training and fine-tuning decisions. The companies that control the most powerful and widespread AI models now have immense influence over people and society. Engineering decisions by Google, OpenAI, Anthropic, Meta, and a few other companies are shaping people's experiences, choices, access, and opportunities. The AI layer is swallowing other institutional and social layers, such as governmental, civic, educational, legal, health, and informational ones. The weight of AI's influence is disproportionate and is making everything else orbit it and alter its behavior accordingly. Letting every institutional layer be ruled by the laws of the digital architectural layer (AI) established by a handful of AI companies is risky and might lead to institutional collapse. It's also a path of no return that might lead to societal chaos. The different forces that regulate society must be rebalanced, and AI's socioeconomic influence must be legally contained. The influence of AI must be no greater than that of the market, law, social norms, and other institutional forces. We do that through public debate, comprehensive AI regulation, strengthened institutions, focus on human values and rules, and by treating AI as what it is: a tool. Not alien intelligence, not a special entity, a tool. Subject to the human full stack of regulation and control. On the topic, read my full manifesto for biological intelligence (link below). - Image: @lessig's pathetic dot theory from his book "Code and Other Laws of Cyberspace."
@BoringBiz_ ·
Dario: “we should model AI regulation on agencies like the FAA. Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety” In other words, entrench Anthropic as the incumbent and don’t allow any new competitors to enter the space I have never liked a product but hated the management team and ethos of a company this much before
@LuizaJarovsky ·
🚨 The stakes are too high in AI, and as someone who has been analyzing AI policy developments for over 3 years, I say that it's time for AI regulation to move on. AI companies should have to prove that their AI models and systems are compliant, safe, and also that they have clear societal benefits. Major AI labs should show what percentage of their revenue is being reinvested in societal good initiatives, like AI to help diagnose and cure diseases. Society shouldn't be forced to play the social media game all over again just because AI CEOs want to become billionaires. It's the first time in history that we have the potential for massive cognitive automation and replacement, and companies should be held accountable, proportionally to the risk their products pose. As I wrote above, it's essential that AI regulation moves to the next stage.
@alex_verem ·
Free University of Bozen-Bolzano, Oxford, University of Tartu, IBM Research, and a dozen other institutions just published the most important framework nobody in the AI agent space is reading. Everyone is building AI agents. Nobody is governing them. The academic world just wrote the manual. > The problem isn't that AI agents don't work. It's that when they fail, break compliance rules, contradict each other, or drift from their original behavior after self-modification, nobody has a systematic way to catch it, explain it, or fix it. The current industry approach is vibes-based governance system prompts, retry logic, and prayer. > This manifesto introduces Agentic Business Process Management as the missing layer. The core idea: every AI agent operating inside an organization needs process awareness a continuous understanding of what the organization is trying to accomplish, what constraints apply, and how its decisions align with or violate those goals. Without this, you don't have an AI system. You have an autonomous entity making consequential decisions inside your company with no frame of reference for what it's supposed to be doing. > The framework identifies four capabilities every governed AI agent must have. Framed autonomy means the agent operates within explicit normative constraints not just "follow these instructions" but a formal specification of what it's permitted, obligated, and prohibited from doing. > Explainability means the agent can articulate why it made each decision, not as a post-hoc justification but as a live, auditable trace. Conversational actionability means it can coordinate with humans and other agents in natural language while actually executing process steps not just talking about them. Self-modification means it can adapt to new conditions and evolve its behavior over time without breaking the frame it operates within. > The liability finding is the one that should stop every enterprise AI team cold. When an agent self-modifies and its behavior drifts significantly from its original design, who is responsible for the outcome? The software developer? The organization that configured the frame? The AI model provider? The multi-agent system that produced emergent behavior nobody anticipated? The EU AI Act is coming. None of the current agentic AI deployments have a clear answer to this question. → Models trained for mathematical domains generalize poorly to legal, medical, and business contexts the domains where governed agents are being deployed → CTGAN debiasing changes the direction of disparity without eliminating it purely algorithmic fixes cannot substitute for structural governance → Framing must be both normative (what the agent is allowed to do) and operational (how it achieves goals within those constraints) → Self-modification creates a responsibility gap no current legal framework adequately addresses → Benchmark contamination is identified as a cross-cutting threat — agents employing models trained on benchmark data will fail in real-world conditions The uncomfortable reality this manifesto exposes: the industry is deploying agentic systems at scale while the governance infrastructure is still theoretical. The gap between "we built an AI agent" and "we can audit, explain, and control what that agent does inside our organization" is enormous and it's widening every quarter.
@realBigBrainAI ·
Sundar Pichai, CEO of Google, on the AI policy fights that actually matter to Google: While much of the AI debate centres on hypothetical risks and far-off scenarios, Sundar says Google is focused on something different. The problems that are already here. "I view our role as engaging with policy makers and regulators to help them think through the areas which I think are much more important in the near term." First on the list is energy. As AI infrastructure demands enormous and growing amounts of power, figuring out how to scale sustainably has become one of Google's most pressing concerns. "How do we responsibly grow and meet our energy needs in a sustainable way? Things like permitting reform, investing in new types of energy." Next is cybersecurity. @sundarpichai flags the new wave of security risks emerging alongside AI as a challenge that requires close collaboration between industry and government. Then comes the human element. As AI reshapes how work gets done, Sundar stresses the urgency of investing in education and reskilling the workforce. "How do you invest in education and reskilling the workforce?" And perhaps the most unsettling challenge of all is the erosion of truth. Sundar raises the threat of deepfakes and the growing difficulty of distinguishing reality from fabrication. "How do you handle deep fakes? How do we make sure we can all understand what reality is?" What's most striking is Sundar's honesty about how far there still is to go. Even with Google's scale and resources, he admits these problems are far from solved. "We haven't by any means addressed all of those areas. So, I think we have focused more as a company on engaging as well as contributing in those areas." The message is clear. The AI policy fights that actually matter to Google are about energy, security, education, and truth. Challenges that are urgent, tangible, and already shaping how this technology impacts the world.
Watch video
@mattpavelle ·
There's an excellent framework for healthcare AI legislation being developed by @AdamMeier20 and the Cicero Institute. They wrote a model bill that gets the fundamentals right: regulate the practice of medicine, not the code. Their AI Medical Services Act creates a new licensed provider type, requires malpractice insurance, mandates transparency, and more. I like this because the safety decisions related to AI in medicine should be made based on open and accurate data. Interestingly, it builds in a regulatory sandbox with a shot clock: 30 days for completeness, 90 days for a decision, etc. I'm glad to see serious people building things like this. https://t.co/sROwdqn7AT
@ColtonOrtolf ·
States passed 33 healthcare AI laws in 2025 while the feds went hands-off. Healthcare AI startups now face 50 different definitions of 'medical advice,' 'clinical decision support,' and 'patient data.' Same AI tool, 50 different compliance frameworks. We've fractured healthcare innovation into state-by-state silos. Big players with legal armies win by default.
@mark_k ·
The FT asks whether @AnthropicAI talked itself into an AI export ban. Their analysis found Anthropic used risk, regulation, and restriction language around 8x more often than OpenAI this year: 5 words per 1,000 vs 0.6. Then Washington barred foreign access to Mythos and Fable, reportedly over concerns that Fable’s safeguards could be bypassed and the model used to identify software vulnerabilities. The irony is hard to miss: spend years telling policymakers your own models are uniquely dangerous, and eventually they may start believing you. A company can’t endlessly market itself as building civilization-scale risk and then act surprised when governments treat its products like strategic weapons.
@sjgadler ·
New post: A "federal framework" for AI often means "block state laws and replace them with nothing." I wrote about two questions to tell if a framework has real substance, or is just 'sparkling anti-regulation.'
@BrookeSingman ·
EXCLUSIVE: White House unveils its first national AI legislative framework, pushes Congress to act 'this year' #AI @MichaelKratsios @mkratsios47 @DavidSacks @WhiteHouse https://t.co/xytmBc3l5L
@sjgadler ·
I dug into whether OpenAI complied with California’s AI safety law (SB 53). There's a deeper issue, though: Why are we still letting AI companies police themselves?
@DavidSKrueger ·
Anti-trust is cited as a reason why AI companies can't do a coordinated pause. I looked into this a tiny bit and it seems pretty clear this is uncharted territory where the law could go either way. A canonical case seems to be National Soc'y of Prof. Engineers v. United States, 435 U.S. 679 (1978), where it was established that you genereally can’t restrict production due to public safety concerns. But there are two reasons I think this shouldn't apply to developing AIs that pose x-risk: 1) The argument is based on a prioritization of consumer welfare over public safety, and assumes anti-competitive restrictions on production are bad for consumers. But if developing an AI system has a good chance of killing all the consumers, that's pretty clearly bad for their welfare. 2) This precedent is about producing an existing products. Developing a new AI is not the same as making another widget. It's a whole new beast. And there are many different AIs that could be developed. Should the law compel companies to build specifically the kinds of AIs that could kill everyone? Obviously not. I am not a lawyer, and welcome pushback. This is just my quick gloss on the anti-trust situation and how I would approach arguing against this legal concern.
@VanberghenEU ·
Today in Brussels: EU Council Moves to Simplify AI Regulation The initiative forms part of the “Omnibus VII” legislative package, designed to reduce regulatory complexity and improve the implementation of the Artificial Intelligence Act across the Union. The package includes two proposed regulations intended to simplify the EU’s digital legislative framework and ensure more coherent and practical application of harmonised AI rules throughout the EU. Main amendments introduced by the Council The Presidency treated the proposal as a priority, and member states broadly supported the direction of the initial proposal from the European Commission, recognising the urgency of ensuring a smoother implementation of the Artificial Intelligence Act. However, the Council’s mandate introduces several important adjustments. First, it adds a new prohibition targeting AI systems used to generate non-consensual sexual or intimate content, as well as child sexual abuse material. The Council also establishes a clear timeline for the postponed application of rules governing high-risk AI systems. Under the new schedule, rules will apply from 2 December 2027 for standalone high-risk AI systems and from 2 August 2028 for high-risk AI systems embedded in products. In addition, providers will again be required to register AI systems in the EU database for high-risk systems, even when they consider their systems exempt from high-risk classification. The Council also restores the strict necessity standard for processing special categories of personal data when used for bias detection and mitigation. The mandate further extends the deadline for establishing national AI regulatory sandboxes to 2 December 2027. It also clarifies the supervisory responsibilities of the AI Office, particularly regarding AI systems built on general-purpose AI models developed by the same provider. Certain areas -such as law enforcement, border management, judicial authorities and financial institutions - remain under national supervisory competence. Finally, the Council calls on the Commission to issue practical guidance to help companies operating high-risk AI systems comply with the AI Act in a way that reduces administrative and compliance burdens, especially where sector-specific EU legislation already applies. Following the Council’s agreement on its negotiating mandate, the Presidency will now begin interinstitutional negotiations with the European Parliament to reach a final agreement on the proposal.
@Hesamation ·
“AI IS THE ONLY INDUSTRY THAT IS LESS REGULATED THAN A SANDWICH SHOP.” - MIT professor US government has two choices: > pull the EU card, regulate, be “safe” > minimum regulations, lead the AI race the Fable 5 ban was already a big move that might not be the only one we see in the future. but Open Source AI and small labs take the bigger hit from regulations while OpenAI and Anthropic, with huge capital, lawyers, and lobbyists, who active push for more regulations, are the ones who benefit from less competition.
Watch video
@PeterHndrsn ·
Since our piece was cited as agreeing with this post, I want to make it clear. I do not agree, and it's not what we wrote. Just because AI outputs are entitled to First Amendment protections does not mean "almost all regulation of AI is ipso facto illegal." Thoughts👇 Yes, it makes some types of regulation harder. I think that's a good thing! The First Amendment should prevent regulation that forces a particular viewpoint on AI outputs. There's too quick a path from no 1A scrutiny of AI to authoritarianism. Yes, there is lots of poorly crafted legislation that doesn't consider 1A principles and gets struck down. But you *can* draft regulation that addresses AI harms while passing 1A scrutiny. Traditional exceptions (e.g., speech integral to criminal conduct) apply just the same. We also have tiers of scrutiny! Commercial speech gets intermediate (not strict) scrutiny under Central Hudson. And regulating the use of AI can be drafted to appropriately navigate these tiers (attaching a handy chart that might be useful from Congress). I encourage folks to read our piece for more on the contours of this question. When I have more time I'll write up something more detailed on my current thinking.
@WesRoth ·
OpenAI and Anthropic are archrivals. But in Washington, they are quietly working toward the same goal. A federal review system for powerful AI models. The proposal would give government evaluators access to qualifying frontier models for up to 30 days before they are released to other trusted partners. And they do not want the rules applied only to themselves. They reportedly want the same capability threshold imposed across the industry. Meta. xAI. And every other company building models powerful enough to trigger the review. Officially, the system would be voluntary. The White House order specifically says it cannot create mandatory licensing or government permission requirements for releasing AI. But voluntary standards have a way of becoming unavoidable. Enterprise customers may demand proof that a model passed government testing. Investors may treat participation as evidence of responsible development. And companies that refuse could be portrayed as national-security risks. This is where safety policy becomes competitive strategy. OpenAI and Anthropic can afford specialized lawyers, security teams, government relationships, and month-long evaluation processes. Smaller labs may not be able to. So both things may be true. The companies want clearer rules before dangerous models are released. And those rules could strengthen the position of the companies already at the top. OpenAI and Anthropic fight over the AI market. But they may agree on who should be allowed to enter it.
@alex_prompter ·
BREAKING: The first legal precedent for AI agent access rights is being written right now. A federal judge ruled that user consent doesn’t equal platform authorization. The 9th Circuit stayed the injunction days later. The legal framework is still forming in real time. What’s actually at stake: whether AI agents can operate across third-party platforms on your behalf, or whether every platform becomes a walled garden that blocks external automation entirely. This isn’t a Perplexity problem. This is an infrastructure problem for anyone building AI workflows that touch multiple services. Every MCP connection. Every n8n automation that pulls data from external platforms. Every AI agent that logs into a service to execute a task. The legal ground underneath all of it is being defined in this case. Two possible futures: → Platforms control agent access through partnerships and approved integrations. AI works, but only where it’s invited. → User-directed AI agents operate freely across the open web. Platforms adapt or get bypassed. The ruling to watch: the 9th Circuit’s permanent decision, expected in the coming months. That’s the one that sets the standard. If you’re building with AI agents, this case is your regulatory environment taking shape.
@JoshuaTLevine ·
Ensuring the federal government can procure AI systems quickly and effectively is a crucial component of AI policy, but has flown under the radar, save one notable example. Today, my colleagues and I @JoinFAI filed a comment in response to GSAR 552.239-7001, the GSA’s proposal for rewriting the contract terms for how every agency buys AI through the government's largest commercial marketplace (the Multiple Award Schedule).
@AI_4_Healthcare ·
Should autonomous clinical AI be licensed like a clinician? This @statnews OPED proposes competency tests, scoped practice, supervised deployment, monitoring, and federal oversight. The AI Governance debate is evolving from model performance to clinical accountability. 🔗 https://t.co/wK6U2Bfjjm #ResponsibleAI
@SwissCognitive ·
#AIwashing is moving from marketing habit to legal risk, as regulators challenge claims on availability, performance, and proof.⚖️ Companies now need clear evidence, transparent communication, and accurate timelines.🤖 https://t.co/N9zUqu0e3n #AI #AINews
@iamKierraD ·
You know the answer. lol AI projects at work. You’re already in legal…if there’s not an AI governance committee or something of the sort, ask to create it. Create an AI governance framework for engineers to use. Risk management assessments. Vendor compliance checklists.
@tonysimons_ ·
Today is a pretty damn important day for AI. The EU AI Act’s transparency rules officially begin applying. That means new disclosure requirements for: - AI systems interacting with people - Deepfakes - Certain AI-generated public-interest content - Machine-readable detection of generated media The real question now: Will platforms make disclosure useful, or bury it behind another meaningless little badge? https://t.co/FSpwavXEFf
@tbpn ·
"Federal preemption is good when you are pro-regulatory capture." Asked about the discussion over AI regulation, @travisk says companies push for broad federal regulation when it helps squeeze out competitors. "When you want to squeeze others out, you should get federal regulatory bigness going for you." "We never did that at Uber. We basically never ever proposed or pushed any rule that would be beneficial to us versus somebody else. We always were trying to open up the market and we said, 'Let the best man win' and we just went for it." "I think we've got to be careful of some of these closed-weight things that are creating situations where they need to be regulated."
Watch video
@shawnchauhan1 ·
OpenAI just published a 13-page document telling the government how to regulate AI. Read that sentence again. The $852B company building the technology is also designing the compliance framework for that technology. It explicitly calls for auditing "a narrow set of highly capable models" - which is exactly what OpenAI builds. It flags biosecurity, drug discovery, and autonomous cybersecurity as high-consequence. Those are the three most actively funded AI niches in venture right now. Framing this as a social contract is smart. What it actually does is raise the floor for every competitor not named OpenAI. The memo is worth reading. The fine print is worth reading more carefully.
@MattPerault ·
Protect kids when they use AI. Let kids use AI. Too many proposals treat these as competing goals. We lay out a framework that embraces both. Don't ban minors from AI. Learning these tools is becoming as essential as learning to use a computer. But require platforms to offer parental controls, prohibit misrepresentations about AI tools, and publish protocols for handling crisis situations. And enforce the law when AI is used to harm kids. New on the a16z AI Policy Brief today.
@sagar_batchu ·
Most companies with an AI governance program have an acceptable use policy that went through legal, got circulated at all-hands, and now lives in the employee handbook. That used to be sufficient when leaking customer PII couldn't happen accidentally from a single AI agent session. In practice, an acceptable use policy only governs what an employee does manually. It does not govern what their agents do on their behalf. An employee knows not to paste customer records into an external tool. An agent often doesn't. Customer security questionnaires are already asking about this distinction. "Do you have an AI acceptable use policy?" is a checkbox. "How do you enforce this?" is the follow-up that most companies cannot answer. Enforcement has to live inside the agent lifecycle, at the point where tool and model calls execute. Not in a skill file your agent forgot to read. Policies govern people. An AI Control Plane governs agents.
@jbobbink ·
Lets talk about a gap in Google's new AI content labeling. They require it for product feeds but only "recommend" it for forums. Google just updated its Discussion Forum and Q&A Page structured data docs with a new property called digitalSourceType. It lets sites flag when a post or comment was written by an AI model or an automated bot. The idea sounds great on paper. In practice, the implementation tells a different story. The property is listed as "recommended," not required. If a site leaves it out, Google assumes the content is human-generated. That is a massive loophole. Every forum operator running AI-generated answers at scale can simply ignore the field and Google will treat that content as if a real person wrote it. Now compare this to Google Merchant Center. There, AI-generated product descriptions must include the digital source type field. It is required, not optional. Same IPTC standard. Same labeling vocabulary. But a completely different enforcement approach depending on where the content lives. Why does this matter for SEO? Forum content is one of the most influential content types in Google Search right now. Reddit threads, Stack Overflow answers, and Quora posts show up in featured snippets, AI Overviews, and "Discussions and forums" carousels. If Google cannot reliably tell which forum responses are AI-generated, it is building its most prominent search features on content it cannot verify. The regulatory world is not waiting around. China mandated AI content labeling in 2025. Spain proposed fines of up to 35 million euros for non-compliance. A March 2026 study by Indicator found that major platforms including LinkedIn and TikTok still have significant gaps in detecting AI-generated content, even when proper metadata exists. Google has the leverage to fix this overnight. They could tie digitalSourceType to rich result eligibility for forums. They could flag pages that omit it. They could do what they did with product feeds and simply require it. The fact that they chose "recommended" instead raises an uncomfortable question. Does Google actually want to know how much forum content is AI-generated? Or would the answer be too inconvenient for a search engine that just built an entire SERP feature around it? Sources: - Matt G. Southern, "Google Adds AI & Bot Labels To Forum, Q&A Structured Data" (March 2026), Search Engine Journal - Google Merchant Center Help, "AI-generated content" (2025), Google Support - IPTC, "AI disclosure on social media a work in progress" (March 2026), https://t.co/yAEsZ7t3QG
@lemire ·
Ideas have consequences. Not all bad ideas end up with mass murders… thankfully… but ideas have consequences nonetheless. Currently, you can just go to your favourite AI (Grok, Gemini, Claude, ChatGPT) and ask anything. In some cases, the vendors themselves censor the AI. Gemini might refuse to answer a political question, for example. Importantly, the AI that helps you code won’t stop you from designing a website that criticizes the immigration policy of your current government. The AI won’t stop you from double-checking the diagnosis your medical doctor gave you. But this could easily change. We have had far-reaching regulations in Europe and elsewhere… and more could easily come. “But what is wrong, Daniel, with ensuring that AI is safe?” The problem is regulatory capture. Imagine that I am a financial advisor sick of seeing my clients double-check my recommendations with ChatGPT. I can simply go to a politician and propose that ChatGPT be held accountable if its financial advice ends up wrong. You just allow citizens to sue OpenAI for bad advice. You are IBM and want to protect your consulting business. You make sure that if someone uses Anthropic’s AI Claude to write COBOL and there is a bug, then Anthropic can be sued. You are a pharmaceutical company that does not like it when an AI recommends against a particular medication due to its side effects? Make sure AI providers can be sued for medical malpractice. You are a politician who does not like what the AI tells your citizens about you? Make sure AI vendors must get a licence from the state to offer their services—and as a condition for maintaining their licence, they must pass stringent compliance tests every few months. Every single one of these steps can then lead to a spiral of regulations that might halt further innovation. What protects us from such a fate? Maybe not as much as we’d like. If you think that your AI coding agents can't possibly be forced to monitor your work in the future... think again. Thus it is important to fight in the realm of ideas and culture right now.
@MattPerault ·
Building an AI startup is hard. Small teams must compete with well-resourced incumbents, fighting for access to talent, data, and compute. But in California, there’s another barrier that increasingly makes it hard for startups to compete: a dense and fast-growing web of regulations. What does a small team of founders need to do to comply with California’s AI regulations? We mapped the compliance obligations AI startups face in California in the wake of new laws passed in California in the last few years. A startup’s success shouldn’t be determined by its legal budget. If startups need to write big checks to lawyers just to get an AI product into the market, we’ll end up with fewer challengers, worse products, and a less dynamic AI ecosystem. Greetings from the Garage: https://t.co/0rK7P3aykd
@mikekatz29 ·
“Federal AI clarity” via Presidential Executive Order (as we saw last month) sounds comforting. It may make founders and builders think they don't need to comply with more aggressive state AI laws and regs. But in reality, it’s the opening move in a coming preemption war between Washington and the states, with years of legal volatility ahead for companies actually building AI. More on this below ⬇️
@nikillinit ·
When AI starts doing clinical tasks, how should malpractice work? The subject of tomorrow's post - there are a few dimensions that AI is significantly changing how we think about what liability means IMO there's room for new malpractice carriers to be built that need different sets of skills (e.g. diligencing clinical agents) and provide different products (e.g. combo malpractice/professional liability/cybersecurity)
@richardokunolar ·
Using AI Responsibly I recently delivered a session on this topic during a KSS meeting at work. I presented two cases where professionals faced legal consequences after relying on AI generated information without fact-checking. AI tools such as Claude and ChatGPT are super fast and good, we know that, but will they take responsibility for errors contained in a client work? Will they take care of the underpaid tax, penalty and interest caused as a result of negligence? In one case the individual was sued, and in another the person was fined after it was discovered that the AI had cited an incorrect case within the advisory opinion. The above examples show that accountability for AI generated output still rests with the professional who signs off the final work. This highlights the importance of using AI responsibly and declaring its use where appropriate. The final output of any AI generated content is signed off as you. Make sure to check the credibility of every data.
@jeditrinupab ·
Every company using AI needs simple rules: 1. What data cannot go into AI tools 2. Which outputs require human review 3. Which tools require company accounts 4. Who owns final approval 5. How AI-generated work is labeled Without rules, AI makes you faster and riskier.
@intern_lm ·
The word "agent" carries two very different meanings. In AI, an agent is a system that can perceive, plan, and act. In law, an agent receives authority but also assumes a duty of loyalty and can be held accountable. Today's AI agents can receive the first of those things. They cannot assume the rest in the way a person or company can. Authority is beginning to flow to machines while responsibility remains in human society. Mark Nitzberg, the Executive Director of the Center for Human-Compatible AI and Head of Strategic Outreach at Berkeley AI Research，described this as a gap between engineering and law. A human agent who fails in a duty can be investigated and sanctioned. A probabilistic black-box system has no engineering equivalent of a fiduciary duty. Closing that gap, he argued, requires AI that is well-founded, legible, and steerable: built from components people can reason about, understandable in operation, and stoppable or changeable at any time. Lan Xue，Dean of Schwarzman College and the Institute for AI International Governance at Tsinghua University，noted that classic principal-agent problems have not vanished. Information asymmetry, moral hazard, and misaligned objectives all persist. What has changed is the accountability chain. A machine cannot be punished, so responsibility must be redistributed across design, development, deployment, and use. Nicholas B. Dirks, President and CEO of The New York Academy of Sciences，warned that delegation is never only about handing off a task. In the pursuit of efficiency, people may also delegate goals, judgment, and principles. "If we are only thinking about AI that way," he said, "we are already beginning to forget about our values." #WAIC2026 #agent
Watch video
@CRSegerie ·
I agree with @deanwball that getting AI child-safety laws right is quite challenging AI law could backfire, yes But my biggest criticism of this is that there are no numbers or a prediction of the amount of illegal usage If 5% of kids would migrate to unregulated Chinese open-weight models while 95% benefit from guardrails on mainstream platforms, the law is overwhelmingly worth passing
@jacobm ·
2026 is the year when Governance Becomes Culture As work becomes increasingly system-run, culture changes form. In the past, culture lived in norms, how people behaved when no one was watching. In an agent-driven environment, behavior is shaped by rules, permissions, defaults, and escalation paths. Culture becomes embedded in code, policy, and process. This isn’t philosophical. It’s operational and regulatory. In Europe, the EU AI Act is already reshaping how organizations deploy AI in employment contexts, with prohibitions on certain practices (such as emotion tracking of workers) and full implementation scheduled for August 2026. In the U.S., New York City’s Local Law 144 requires bias audits and transparency for automated employment decision tools. Globally, standards like ISO/IEC 42001 are emerging to formalize AI governance. These frameworks aren’t about innovation theater. They’re about legibility: being able to explain what a system did, why it did it, and who is accountable. What this means in practice: When an agent flags an employee, reroutes work, or influences a promotion decision, the organization must answer basic questions: Who approved this? On what data? With what recourse? Culture is no longer what leaders say it’s what systems allow.
@GlenGilmore ·
AI Governance Lessons Businesses Can't Ignore In 2026 “EU member states designated enforcement authorities in August 2025, with the penalty regime now in effect—fines reaching €35 million or 7% of global revenue for the most serious violations. In the U.S., states are building their own patchwork of requirements—260 AI-related measures were introduced across 47 states in 2025, with 22 already passed into law.” https://t.co/sxhNG8dVTT
@GlenGilmore ·
AI Enforcement Accelerates as Federal Policy Stalls and States Step In Federal agencies are relying on existing authorities to regulate AI-related conduct: FTC: Section 5 of the FTC Act remains a primary enforcement vehicle for allegedly unfair or deceptive AI practices SEC: focused on so-called “AI washing” misrepresenting the use or performance of AI in disclosures to investors False Claims Act: DOJ has signaled willingness to pursue FCA theories where AI tools are used in government-funded programs, including healthcare reimbursement and cybersecurity compliance contexts. Antitrust Enforcers: DOJ and FTC https://t.co/stZVb48cD2 #AIGovernance
@theinformation ·
How is the government approaching AI oversight? The Information's Leo Schwartz explains: "We just have this new normal, which is essentially a de facto licensing regime with no rules." "There was a whole clause that said nothing in here can be construed to be mandatory or to be a licensing regime."
Watch video
@arrotu ·
Most AI startups are building for performance. The bigger opportunity in 2026 may be building for auditability. Under the EU AI Act, high-risk systems in lending, insurance, hiring, education, and similar areas will need stronger traceability and defensibility. I wrote about the kinds of high-risk AI businesses that could win — and why verifiable execution may become core infrastructure for them. https://t.co/xQMjtPmZtC via @LinkedIn
@CRSegerie ·
What's missing in AI policy right now: a piece that tells the White House what to do post-Mythos, written in its own values, structuring the ad-hoc prior restraint with clear go/no-go criteria. Hawks: your move.
@1DavidClarke ·
Everyone is talking about AI governance. But here's the uncomfortable truth. In most organisations, AI is already a GDPR issue. Why? Because the majority of AI systems process personal data — whether in the input, the processing, or the output. The model may analyse behaviour. The output may profile, score, rank, or influence decisions about people. Not every AI system triggers GDPR. But far more do than most organisations realise. Look at what organisations are already deploying: • customer service copilots • recruitment screening tools • fraud detection systems • marketing recommendation engines • credit and risk scoring models All of these involve profiling or analysing identifiable individuals. Under GDPR, that is personal data processing. So before we get lost in complex AI frameworks, the first question should always be simple: Does this AI system process or affect personal data? If the answer is yes, the organisation must address the core GDPR obligations. In practice, the key areas to focus on are: Lawful basis — Why is personal data being used in the AI system? Purpose limitation — Is existing data being reused for AI training or analytics beyond the original purpose? Data minimisation — How much personal data is actually going in, and is all of it necessary? Transparency — Do individuals know AI is involved in the decision or interaction? Profiling and automated decisions (Article 22) — Is the AI making or influencing decisions about people? DPIAs — High-risk AI systems should trigger data protection impact assessments. Accuracy and fairness — Incorrect AI outputs about individuals can create GDPR risk. Human oversight — Can a human review or override AI outcomes? These aren't theoretical. They're live compliance obligations. Yes, broader AI governance matters. But for most organisations, the biggest immediate exposure sits somewhere much more familiar. Their AI is simply another form of personal data processing. Start with GDPR. Then build AI governance on top of it. What's the biggest GDPR blind spot you've encountered (or recently fixed) in your organisation's AI projects?
@TheSixFiveMedia ·
Should AI be regulated like a utility? @DanielNewmanUV and @PatrickMoorhead debate it this week on The Flip. One side: AI is becoming infrastructure, with real power and access constraints The other: regulation kills innovation and risks handing leadership to China “Utilities are designed for stable tech. AI is doubling every 12 months.” So what happens if we treat it like one?
Watch video
@BernardMarr ·
Could Your AI Content Land Your Business In Court? #GenerativeAI is helping businesses create content faster than ever, but it is also creating new #copyright and #intellectualproperty #risks that many companies have not fully understood. As lawsuits increasingly focus on how AI-generated #images, #videos and #marketing materials are used in the real world, leaders need clearer policies, stronger review processes and better safeguards before publishing #AI #content at scale. https://t.co/yuERtk1gtg
Best AI Regulation tweets
Xholic studies what works in your niche, drafts posts in your voice and schedules them for the hours your audience is online.
$0 today · Cancel anytime
Browse all tweet collectionsKeep exploring