<\/p>\n
Three developments last week underscored how fragmented AI governance has become: CNN sued Perplexity AI over alleged copying of its journalism; OpenAI published a governance framework tied to emerging EU and California rules; and the legal battle over state AI regulation in the US continued to intensify.<\/p>\n
The disputes show AI policy evolving on multiple fronts at once. Copyright enforcement, safety disclosure requirements and constitutional fights over state regulation are advancing separately, forcing AI companies to navigate overlapping legal and political risks rather than one coherent regulatory system.<\/p>\n
CNN\u2019s lawsuit, filed in federal court in New York, accuses the AI search company of scraping and republishing more than 17,000 news stories, photographs and videos without permission. The broadcaster says Perplexity infringed copyright and misused its trademark by implying a content relationship that did not exist. Perplexity\u2019s response was familiar. \u201cYou can’t copyright facts,\u201d said Jesse Dwyer, the company\u2019s chief communications officer.<\/p>\n
The case arrives amid growing resistance from publishers to AI firms using journalism to power search and answer tools. Perplexity already faces claims from several media groups, including The New York Times, News Corp and Dow Jones, while outlets such as Time, Gannett, Le Monde and Der Spiegel have signed licensing deals instead. The CNN case adds to mounting pressure on AI companies over how they source and commercialise protected material.<\/p>\n
OpenAI, meanwhile, moved in a different direction. The company published its Frontier Governance Framework, setting out how its internal safety systems align with the California Transparency in Frontier AI Act and the EU AI Act\u2019s Code of Practice for general-purpose AI.<\/p>\n
The framework addresses risks including cyber abuse, chemical, biological, radiological and nuclear threats, manipulation and loss of control. OpenAI said it will update the document as regulation evolves.<\/p>\n
The publication reflects a broader shift towards pre-emptive disclosure in frontier AI. California\u2019s law took effect on 1 January 2026, while the EU\u2019s transparency rules become enforceable on 2 August 2026. OpenAI\u2019s framework gives regulators, rivals and customers a public benchmark for how the company says it manages high-risk model behaviour.<\/p>\n
Alongside those developments, the fight over whether US states can regulate AI independently continues. In Colorado, xAI has challenged the state\u2019s AI law in federal court, with the Department of Justice intervening in support of the company. Lawmakers have since signed a narrower replacement statute, but the litigation has left enforcement uncertain and exposed the difficulties companies face operating across multiple jurisdictions.<\/p>\n
Taken together, the three episodes suggest AI governance is not converging around a single model. Copyright disputes, state-federal legal conflicts and frontier safety regimes are all advancing simultaneously, each with different rules, deadlines and costs. For AI companies, compliance is becoming less a unified strategy than a series of separate legal and political battles.<\/p>\n