★ Top Story
The VergeOn June 12th, the Trump administration issued an export control directive ordering Anthropic to suspend foreign access to its newly released Mythos 5 and Fable 5 models — models Anthropic described as exceeding any previously released AI in capability. The order forced the company to take its most powerful models offline over the weekend, affecting even its own non-US employees. The incident has sparked a broader debate about government overreach into the AI industry and the fragility of US AI dominance.
Read the original →
The VergeThe sudden US-mandated takedown of Anthropic's frontier models has sent a stark signal to the rest of the world: American AI can be switched off by Washington at any time. Foreign governments and enterprises are now reconsidering their dependence on US AI providers, accelerating interest in sovereign AI alternatives. The episode may prove to be a turning point in the global AI landscape.
Read more →
TechCrunchDozens of cybersecurity professionals have signed an open letter urging the White House to lift export-control restrictions on Anthropic's Fable 5 and Mythos 5 models, warning the ban undermines defenders more than adversaries. The experts argue that restricting access to state-of-the-art AI tools hampers the ability of security teams to detect and respond to threats. The protest adds to growing industry pressure on the administration to reverse the directive.
Read more →
TechCrunchSpending data from corporate card platform Ramp suggests that Anthropic's enterprise adoption has continued to accelerate even as it battles the White House over export controls. The company's positioning as a principled actor willing to push back against government overreach appears to be resonating with business customers. The dynamic is an unusual one: political conflict may be functioning as a brand differentiator.
Read more →
Ars TechnicaJust days after its blockbuster IPO, SpaceX has announced a $60 billion all-stock deal to acquire Cursor, the popular AI-powered coding assistant. The acquisition is a direct bid to close the gap with Anthropic and OpenAI in the enterprise AI market, where SpaceX's existing AI division has struggled to gain traction. SpaceX told IPO investors it sees a $26 trillion addressable market in AI, and the Cursor deal is its most aggressive move yet to claim a share of it.
Read more →
Ars TechnicaAudited financial documents leaked from OpenAI reveal that the company's rapidly growing revenues are being far outpaced by research, development, and operational expenses, resulting in multi-billion-dollar annual losses. The disclosure raises fresh questions about the long-term financial sustainability of frontier AI development and the timeline to profitability. The documents come at a sensitive moment as OpenAI continues its ongoing restructuring into a for-profit entity.
Read more →
TechCrunchChatGPT has fallen below 50% market share among AI assistants for the first time, though it remains the dominant player with over 1.1 billion monthly active users. Google's Gemini has surged to 662 million users, while Anthropic's Claude has reached 245 million, reflecting a rapidly fragmenting competitive landscape. The milestone signals that the era of ChatGPT's near-monopoly on consumer AI is definitively over.
Read more →
Ars TechnicaThe Justice Department has intervened in an NAACP lawsuit against xAI, arguing that the company's unpermitted gas turbines powering its Grok data center constitute a matter of "national, economic, and energy security" because the Pentagon relies on the Grok AI system. The move is a remarkable instance of the federal government using national security arguments to shield a private AI company from environmental enforcement. Critics say it sets a dangerous precedent for sidestepping pollution regulations in the name of AI infrastructure.
Read more →
Ars TechnicaAnthropic has reversed course on a planned billing change to its Claude Agent SDK that would have dramatically increased costs for heavy users by switching to token-based pricing. The move, originally set to take effect Monday, drew significant backlash from developers who rely on the SDK for agentic workflows. The pause suggests Anthropic is still calibrating how to monetize its developer ecosystem without alienating its most active builders.
Read more →
Ars TechnicaThe Department of Defense has publicly touted its use of generative AI to draft official reports required by Congress, claiming that 1.5 million military personnel are now using AI tools. The disclosure raises accountability questions about whether AI-generated government documents meet the standards of accuracy and oversight that congressional oversight demands. The announcement comes amid broader scrutiny of AI's expanding role in military decision-making.
Read more →
NVIDIA BlogNVIDIA's Blackwell GPU architecture has dominated the latest MLPerf Training 6.0 benchmarks, setting new records across speed, scale, and reliability metrics for AI model training. The results reinforce Blackwell's position as the de facto standard for frontier AI training infrastructure at a time when model scale and training iteration speed are critical competitive advantages. The sweep comes as NVIDIA also moves to raise over $25 billion in its first bond offering since 2021.
Read more →
Ars TechnicaNVIDIA is tapping debt markets for the first time in five years, seeking to raise more than $25 billion in a bond offering that will test investor appetite for continued AI sector exposure. The capital raise comes as the company faces surging demand for its Blackwell chips while simultaneously investing heavily in next-generation architectures. The deal is one of the largest corporate bond offerings in the tech sector in recent memory.
Read more →
TechCrunchGoogle has officially released Android 17 alongside Wear OS 7, bringing a suite of new multitasking capabilities, enhanced parental controls, and deeper Gemini AI integration across its device ecosystem. A simultaneous Pixel Drop pushes Google's latest AI models directly to Pixel hardware, tightening the link between Google's software and AI ambitions. The release marks one of the most AI-centric Android updates to date.
Read more →
Ars TechnicaA newly disclosed vulnerability dubbed "SearchLeak" in Microsoft Copilot allowed attackers to exfiltrate two-factor authentication codes from users, highlighting persistent security weaknesses in LLM-powered enterprise tools. The exploit is the latest in a string of AI assistant security failures that researchers say stem from fundamental flaws in how the industry approaches LLM security architecture. Microsoft has patched the issue, but the incident renews calls for more rigorous security standards around AI integrations.
Read more →
TechCrunchSalesforce has agreed to acquire Fin, an AI-native customer service platform, for $3.6 billion, with plans to integrate Fin's technology and team into its Agentforce enterprise AI product. The deal reflects Salesforce's urgency to bolster its agentic AI capabilities as competition in enterprise automation intensifies. It is one of the largest AI acquisitions in the CRM space to date.
Read more →
TechCrunchMeta has launched an "AI Mode" search feature on Facebook that generates answers by drawing on public posts and content from across Meta's family of apps. The rollout is part of a broader wave of new AI features Meta is deploying as it works to close the gap with rivals and keep users more deeply engaged on its platforms. Privacy advocates are likely to scrutinize how user-generated content is being used to train and power the AI responses.
Read more →
MIT Technology ReviewCasey Harrell, who has ALS and is fully paralyzed, has logged thousands of hours using a brain-computer interface (BCI) that translates his neural signals into speech — making him the most experienced long-term BCI user in the world. His case offers the most detailed real-world evidence yet of how BCIs perform over years of continuous use, rather than in controlled research settings. The milestone represents a significant step toward BCIs becoming practical, everyday assistive devices.
Read more →
TechCrunchA new survey from WordPress VIP finds that 60% of US consumers react negatively when brands prominently feature "AI" in their marketing, even as companies increasingly invest in AI-powered products and AI search as a customer acquisition channel. The gap between corporate enthusiasm for AI branding and consumer skepticism poses a real challenge for marketers trying to communicate product value. The findings suggest that how AI is talked about may matter as much as what it actually does.
Read more →No stories from this source today.