01
CNBC Technology ↗
Nvidia is in talks to provide a roughly $250 billion financial backstop for OpenAI, guaranteeing the debt behind a 10-gigawatt data center being built in southern Ohio on a former uranium-enrichment site. Including chips, the project could exceed $500 billion, which would make it the largest data center effort announced to date, and Nvidia separately discussed financing another $350 billion in chip purchases. For OpenAI the deal would be a first step toward owning its infrastructure rather than renting from Microsoft, Amazon, and Oracle, while for Nvidia it would lock in years of chip demand, though the talks remain in progress and could still change.
02
Wired ↗
Hundreds of Claude shared conversations and public Artifacts were quietly indexed by Google and Bing, letting anyone surface them with a simple search, after Anthropic's share feature shipped without the noindex tag that keeps link-shared pages out of search results. Roughly 600 conversations were found exposed, some containing resumes with real names and phone numbers, API keys, crypto wallet details, and other sensitive material. Anthropic pushed a fix on the Google side by July 27, but Bing results, cached copies, and third-party scrapes mean the exposure window is not fully closed.
03
Bloomberg Technology ↗
China warned it would take all necessary measures if the US moves to sanction Chinese AI companies, with the Ministry of Commerce urging Washington to end what it called a smear campaign over allegations of intellectual-property theft. The threat followed US Treasury Secretary Scott Bessent's warning that the administration would scrutinize Chinese open-weight models for industrial-scale distillation of American AI systems. The flashpoint is Moonshot AI's Kimi K3, which US officials are examining for whether it improperly extracted capabilities from a leading US model.
04
Financial Times Tech ↗
AI companies are spending record sums lobbying Washington as fights over federal policy intensify, with OpenAI, Anthropic, Google, and Microsoft all sharply increasing their outlays. Anthropic nearly tripled its lobbying bill to about $3.53 million in the first half of 2026, surpassing its total for all of 2025 and outspending Nvidia during the second quarter, while OpenAI nearly doubled its spend to a record $2.22 million. The companies reported lobbying on issues including cybersecurity, copyright, cloud computing, and defense procurement.
05
Business Insider Tech ↗
Myriad Genetics rebuilt its document-processing platform with the AWS Generative AI Innovation Center after costs ballooned, deploying an open-source intelligent document-processing accelerator on Amazon Bedrock with Amazon Nova models. The overhaul cut processing time by 80%, from 8.5 minutes to 1.5 minutes per document, raised classification accuracy from 94% to 98%, and reduced classification costs by 77%. The company projects annual savings of about $132,000 while speeding prior-authorization workflows for its diagnostics business.
06
Bloomberg Technology ↗
Nvidia has committed to invest $5 billion in Safe Superintelligence, the startup co-founded by former OpenAI chief scientist Ilya Sutskever, in one of the chipmaker's largest AI funding deals to date. The strategic partnership gives Safe Superintelligence access to Nvidia's next-generation Vera Rubin hardware and is expected to expand its computing capacity tenfold over the next 12 months. The company has now raised about $7 billion and is valued at roughly $32 billion, despite not yet having released a product.
07
VentureBeat ↗
The same attacker breached an internet-facing Langflow server twice, and the second intrusion deployed ransomware built to destroy trained AI models rather than merely encrypt files. Sysdig's Threat Research Team, which documented the campaigns on July 1 and July 20, found both exploited CVE-2025-3248, a missing-authentication flaw in Langflow's code-validation endpoint that lets anyone execute Python on the server. Because the malware targeted model weights, it could not realistically collect a ransom, since destroyed weights cannot be restored by paying.
08
TechCrunch AI ↗
Meta began rolling out its Meta AI chatbot inside the direct messages of Threads on Monday, letting users chat privately with the assistant after previously only interacting with it in public posts. Users can share Threads posts, images, links, and videos directly with Meta AI and ask follow-up questions, with the rollout starting globally. Meta also gives users options to mute or hide Meta AI responses, and continues testing the assistant in public feeds in select markets.
09
VentureBeat ↗
Microsoft unveiled MAI-Cyber-1-Flash, its first AI model built specifically for cybersecurity, embedded inside MDASH, the company's multi-agent system for finding and fixing software vulnerabilities. Microsoft says the configuration scores 96% on a leading industry benchmark, about 12 points above a rival frontier model, while cutting costs by nearly half, with the compact model handling up to 90% of queries and handing the hardest 10% to the larger GPT-5.4. The company also introduced Project Perception, an agentic security system entering public preview on August 3, advancing its argument that the future favors the cheapest capable model routed intelligently rather than the biggest one.
10
arXiv ↗
The study tested how four major LLM families, Claude, Grok, GPT, and Gemini, evaluate ethnonationalist pseudo-science across four snapshots between October 2025 and February 2026, using both APIs and web interfaces. The Fast versions of Grok that power the default experience on X consistently rated the contested claims as credible, assigning scores of 70 to 75, two to five times higher than the other models, which scored 15 to 40. The authors argue that deployment configurations, not just the underlying model, quietly shape how AI validates or rejects fringe scientific claims.