Topic
Export controls, antitrust, standards bodies, right to repair and privacy regulation.
Federal banking agencies are not writing a single AI rulebook. They are applying existing model-risk, fair-lending, anti-money-laundering and third-party oversight frameworks to a technology that now touches credit decisions, fraud systems and customer service. Here is what that scrutiny actually looks like.
Persistent memory turns an AI assistant into a long-lived record of what you asked, confessed and planned. The privacy questions are only now getting answers.
The rules restricting advanced AI chips to China are written. The fight has moved to smuggling networks, shell companies, cloud loopholes and the question of whether the controls can actually work.
Dozens of state AI laws are now on the books — on deepfakes, hiring algorithms, disclosure and safety — while federal efforts to preempt them have repeatedly stalled. The result is a patchwork that companies hate and compliance teams are building for.
The constraint on new AI capacity has shifted from chips to electricity: interconnection queues, transformer lead times and utility rate cases are deciding where the next generation of data centers gets built.
Washington treats open model weights as an export-promotion asset; Brussels treats them as a liability question. The same download is now two different legal objects.
The BEAD program was signed in 2021 to connect every unserved American home. Five years on, the money is moving, the rules have changed midstream, and the first shovels are hitting the ground.
Governments have always been the biggest buyers of software vulnerabilities. Now they're starting to regulate the trade — including their own participation in it.
The App Store's rules were set in Cupertino for fifteen years. They are now increasingly set in Brussels, Tokyo, Seoul and Washington — and the services margin story is changing with them.
Direct-to-cell satellites now carry real text messages, and data is next. The dead zone — the last unconnected geography of the mobile era — is being engineered out of existence.
Europe's landmark competition law forced Apple, Google, Meta and others to open their platforms. The compliance is real, the fines are real, and the competitive effects are still hard to find.
The October 2022 rules were meant to freeze China's AI hardware capabilities; three years on, the controls have raised costs and slowed deployment, but adaptation — stockpiles, workarounds and domestic champions — has blunted the freeze.
A wave of US state laws and international rules now require platforms to verify users' ages, and the compliance machinery being built — ID uploads, face estimation, platform-level attestation — will outlast the laws that created it.
With no federal privacy statute in sight, a growing patchwork of state comprehensive privacy laws — each with its own thresholds, rights and enforcement — has become the de facto US privacy regime.
The European Union's AI Act obligations for general-purpose models took effect in August 2025, and their documentation, copyright and risk-assessment duties are quietly becoming the global template — because Brussels is a market too big to route around.
The Federal Communications Commission is set to vote on a satellite spectrum abundance order freeing up bands in the 12 GHz and 42 GHz ranges, alongside rules easing environmental review timelines and modernizing ultra-wideband spectrum for AI-based sensing.
Since August 2, 2026 the European Commission's AI Office has held the power to fine providers of general-purpose AI models up to 3% of global turnover, moving the EU AI Act from a compliance deadline on paper to an enforcement regime with teeth.
Facing surging electricity demand from AI data centers, federal regulators ordered the nation's six regional grid operators to justify or overhaul their large-load interconnection rules within 60 days, aiming to speed access to power without shifting costs onto ordinary ratepayers.
Governor Gavin Newsom signed a package of bills including Adam's Law, requiring AI companion chatbot operators to detect and respond to signs of self-harm in child users, alongside new limits on addictive social media features for minors.
Rulings in Bartz v. Anthropic and Kadrey v. Meta gave AI developers a major win on the fair-use question for training data, while leaving the door open to liability over pirated sources — a split outcome the industry and authors are both still fighting over.
The Trump administration converted roughly 8.9 billion dollars in previously awarded CHIPS Act and Secure Enclave grants into an equity stake in Intel, an unusual move that has drawn scrutiny over how it changes the government's relationship with a company it also regulates and funds.
The Bureau of Industry and Security scrapped the Biden-era AI diffusion framework, then spent 2026 issuing guidance clarifying which older license requirements for advanced chips still apply, leaving exporters navigating a patchwork of rules.
After a DHS funding lapse delayed stakeholder talks, CISA says it still expects to finalize the CIRCIA rule in September 2026, requiring critical infrastructure operators to report major cyber incidents and ransomware payments within tight deadlines.
Commerce Secretary Howard Lutnick said the Trump administration plans to move quickly on approving AI chip export licenses for allied countries, while ruling out any relaxation of export controls on China in the president's upcoming meeting with Xi Jinping.
The European Commission's first Digital Markets Act penalty decisions target Apple's App Store anti-steering rules and Meta's pay-or-consent advertising model, testing how far the bloc's new gatekeeper regime reaches.
The Federal Trade Commission and a bipartisan coalition of state attorneys general allege Amazon used an invented "phantom bidder" to secretly inflate prices in its search advertising auctions, costing advertisers billions of dollars.
Judge Amit Mehta's September 2025 remedies ruling requires Google to share certain search index and click data with rivals while rejecting the Justice Department's push to force a Chrome or Android divestiture.
The Netherlands' data protection authority fined Uber nearly 825 million euros, the second-largest GDPR penalty ever issued, for using fully automated systems to suspend drivers' accounts without adequate human review or explanation.
Performance thresholds written into trade rules now influence how chips are designed.
Software locks can reject a genuine component, which turns a repairable device into a disposable one.
