OpenAI has recently filed a countersuit against Elon Musk in California, seeking an injunction to stop what it describes as "unlawful and unfair action" that has harmed its mission. The legal filing accuses Musk of a fake takeover bid, the near-$100 billion one put forward in February, intended to disrupt OpenAI's transition to a public benefit corporation by the end of the year. OpenAI says that Musk's "continued attacks" threaten to "disrupt OpenAI's future," the company's ability to serve the public interest, and advance safe AI.
During its Google Cloud Next 2025 event, the tech giant launched its 7th-generation TPU named Ironwood, a chip aimed at speeding up inference computing for AI applications like Gemini. The TPU, designed for grouping in configurations of up to 9,216 units, comes with six times larger memory capacity and double the performance of last year's Trillium chip for the amount of energy, making it the company's most energy-efficient TPU yet. Ironwood is part of Google's long-term strategy to reduce reliance on Nvidia's GPUs in pursuit of its own blend of compute in a scalable, cost-effective way.
On Wednesday, The European Commission announced its "AI Continent Action Plan" in a press release to transform Europe's industries and talent "into powerful engines of AI innovation and acceleration." The plan includes building specialized labs and investing €20 billion into five AI gigafactories across the EU, along with streamlining compliance for regional firms via a new AI Act Service Desk. It also mirrors the UK's recent AI expansion efforts and seeks to iron out any remaining legal uncertainties, helping both developers and startups across Europe, the US, and beyond.
According to insiders well-versed in OpenAI's testing processes, OpenAI has significantly cut down its safety testing phase for powerful AI models from several months to just mere days. Critics and former OpenAI staff warn that this rapid timeline of conducting safety checks at earlier "checkpoints" instead of the end product increases risks by potentially missing some "dangerous capabilities." Despite the company assuring that automated testing processes would weed out any catastrophic risks, this news comes at a time when AI safety concerns are at an all-time high.
Following a high-profile dinner at Mar-a-Lago last week, the Trump administration has seemingly backed off its original plans for stricter export controls on Nvidia's H20 chips, which are crucial for AI applications in China. According to one source, the administration was swayed by Nvidia's promise of new investments in US data centers. Despite the continuous pressure from some lawmakers to restrict advanced chip exports, White House officials are delaying further crackdowns, leaving Chinese firms to continue stockpiling these chips.
Just weeks after the first signs for a new Claude plan began appearing, Anthropic announced its Claude Max, a new subscription tier with options starting from $100/month and up to $200/month, much the same as OpenAI's Pro plan. With Anthropic struggling to handle widespread demand and users criticizing Claude's low usage limits, Claude Max offers expanded limits and priority access during peak hours for those willing to pay the cost. The new Claude Max tier is up and running already in all regions where Claude is available.
President Donald Trump is reportedly expected to sign executive orders on Tuesday to give the coal industry a leg-up amid ever-rising power demand for AI data centers, electric vehicles, and crypto mining. The orders look to bring back coal plants that are on the brink of collapsing, reclassify coal as a "critical mineral" for steel production, and end a moratorium on federal coal leasing. With the rise of natural gas and renewable energy sources, the move comes at a time when US coal use has significantly decreased, with backers claiming that existing US coal plants could only provide power to the grid about 40% of the time.
In a report by Business Insider, former OpenAI CTO Mira Murati's AI startup Thinking Machines Lab has reportedly doubled its original seed round goal from $1 billion to $2 billion, which would make it one of history's largest. Despite only recently emerging from stealth and having no product or revenue whatsoever, the round would put the company at a valuation of "at least" $10 billion. The startup's goal remains to, eventually, create AI-based systems that are "more widely understood, customizable, and generally capable" than those of the present day.