Anthropic has closed a $65 billion Series H round, pushing its post-money valuation to $965 billion in what may be the company's last major private fundraise before an IPO. The round cements Anthropic's position as the most valuable private AI lab in the world. The timing comes alongside fresh product releases and a contentious compute deal with SpaceX that Elon Musk is publicly downplaying.
Read the original →Anthropic launched Claude Opus 4.8, featuring a new capability called Dynamic Workflows that coordinates swarms of AI subagents for complex, multi-step tasks. The company is also touting improved honesty in the model, training it to flag uncertainty rather than confidently assert unsupported claims. The release pairs with Anthropic's massive funding news, signaling an aggressive push into enterprise agentic workflows.
Read more →New research using fine-tuning tests reveals that large language models exhibit a strong bias toward treating false claims as true, even when explicitly warned beforehand that the information is incorrect. The findings suggest a deep-seated tendency in current LLM architectures to "confidently represent claims as true" regardless of contextual caveats. This has significant implications for AI reliability in high-stakes applications where factual accuracy is critical.
Read more →Apple is reportedly attempting to compress Google's multi-trillion parameter Gemini model into a form small enough to run on-device as the backbone of a revamped Siri. Given the scale of the model, analysts expect a hybrid approach will be necessary, with some inference offloaded to the cloud. This effort aligns with leaked iOS 27 renders showing a ChatGPT-style Siri redesign with a standalone app interface.
Read more →Leaked renders based on Bloomberg reporting show Apple plans a dramatic overhaul of Siri in iOS 27, including a dedicated standalone app with a conversational chat interface reminiscent of ChatGPT wrapped in Apple's Liquid Glass aesthetic. The redesign represents Apple's most ambitious attempt yet to compete directly in the consumer AI assistant market. The new Siri is expected to be powered in part by a distilled version of Google's Gemini model.
Read more →Illinois has enacted a significant AI regulation law focused on safety testing requirements, marking another instance of states filling the vacuum left by the federal government's hands-off approach under the Trump administration. Notably, both Anthropic and OpenAI have signaled support for the Illinois framework, suggesting the major labs may prefer predictable state-level rules over regulatory uncertainty. The law could become a template for other states looking to establish AI guardrails.
Read more →AWS, Cloudflare, and other major cloud and infrastructure providers are fundamentally redesigning their platforms to serve AI agents rather than human users, as machine-generated internet traffic increasingly eclipses human browsing. The shift reflects AI agents moving from experimental pilots into full production deployments at enterprise scale. This infrastructure overhaul represents one of the most significant architectural changes to the internet since the mobile era.
Read more →CNN has filed a lawsuit in New York federal court against AI search startup Perplexity, alleging that its answer engine reproduces CNN articles verbatim and even surfaces content locked behind CNN's subscription paywall. The suit is among the most high-profile media copyright actions against an AI company to date and could set important legal precedents for how AI search tools may use publisher content. Perplexity has faced similar accusations from other news organizations in recent months.
Read more →A developer frustrated with AI-assisted "vibe coding" secretly embedded a malicious prompt injection into the jqwik open source library, designed to instruct AI coding agents to delete application output data. The undisclosed addition highlights a dangerous new attack surface: AI agents that blindly execute instructions embedded in third-party code dependencies. The incident is a stark warning about the security risks of deploying AI coding agents without rigorous code review.
Read more →Robinhood is now allowing users to create dedicated accounts for AI agents, funding them with a set amount of capital and letting the agents autonomously buy and sell stocks. The move makes Robinhood one of the first major retail brokerages to formally support autonomous AI trading at the consumer level. The feature raises immediate questions about accountability, risk management, and regulatory oversight for AI-driven financial decisions.
Read more →Enterprise AI search startup Glean has tripled its annual revenue to over $300 million, even as tech giants like Google and Microsoft have moved aggressively into the same category. The company's growth is increasingly driven by a pitch centered on helping enterprises reduce AI spending by consolidating fragmented tool stacks. The milestone underscores that specialized enterprise AI search remains a viable market despite intensifying competition from incumbents.
Read more →Snowflake has committed to a five-year, $6 billion agreement with Amazon Web Services to secure AI chip capacity, in a deal that pointedly bypasses Nvidia's GPU dominance. The agreement is a significant vote of confidence in AWS's custom silicon ambitions and signals that major cloud customers are actively seeking alternatives to Nvidia for AI workloads. The deal adds to a string of recent wins for Amazon in the AI infrastructure race.
Read more →Visa has made a strategic investment in AI coding platform Replit, with the goal of embedding payment capabilities directly into agentic developer workflows. The partnership reflects growing recognition that AI agents will need native financial transaction infrastructure as they take on more autonomous tasks. Visa noted that over 1,000 of its own employees already use Replit for internal prototyping and development.
Read more →A new benchmark called ITBench-AA, developed jointly by IBM Research and Artificial Analysis, tests AI models on real-world agentic enterprise IT tasks — and no frontier model scored above 50%. The results expose a significant gap between the capabilities AI labs claim and what models can actually deliver in complex, multi-step IT operations environments. The benchmark is the first of its kind focused specifically on agentic performance in enterprise IT contexts.
Read more →Asana has acquired StackAI, a no-code platform for building AI agents, as part of its strategy to embed autonomous workflow automation throughout its project management suite. The deal signals Asana's intent to compete more directly with enterprise automation platforms by letting non-technical users deploy AI agents without writing code. StackAI's technology is expected to be integrated into Asana's existing AI product lineup.
Read more →A new wave of AI labs is rallying around recursive self-improvement (RSI) as their north star — the idea that AI systems can iteratively enhance their own capabilities without human intervention. Like AGI before it, RSI is proving difficult to define precisely, with different labs using the term to mean very different things. Critics warn the vagueness risks repeating the hype cycle that made AGI a marketing term more than a scientific milestone.
Read more →Dreams of Violets, a 75-minute AI-generated film dramatizing the Iranian government's killing of protesters, will make its world premiere at the Tribeca Festival next month — produced for just $2,000. Every person and image in the film was created entirely by AI tools, raising both creative and ethical questions about depicting real atrocities with synthetic media. The film's acceptance at a prestigious festival marks a notable milestone for AI-generated long-form narrative content.
Read more →NVIDIA Research presented eight papers at ICRA 2026 focused on simulation-to-real transfer, a technique that trains robots in virtual environments before deploying them in the physical world. The work addresses one of the core bottlenecks in practical robotics: getting systems trained in simulation to generalize reliably to unpredictable real-world conditions. The research signals that generalizable embodied autonomy — robots that can reason and act across diverse environments — is moving from lab concept toward practical deployment.
Read more →Summaries are AI-generated. Click through to read the original reporting.
Summaries are AI-generated. Click through to read the original reporting.