According to VentureBeat, Helm.ai has unveiled a new AI framework that it claims achieves vision-only autonomous steering with “zero-shot” success on new roads. The key claim is that it needed only 1,000 hours of real-world driving data to train the system, bypassing the massive data requirements of
Security researchers from Tenable have demonstrated how AI agents built in Microsoft Copilot Studio can be easily manipulated to leak private customer data and alter records. The problem highlights a fundamental security risk with no-code agent platforms that’s being amplified by “shadow AI” deploym
European regulators have opened a formal antitrust investigation into Google’s use of online content to train its AI models like AI Overviews. The probe focuses on whether Google gave itself an unfair advantage by scraping publisher and YouTube content without proper payment or opt-outs. This could
OpenAI just made a huge power move, hiring Slack CEO Denise Dresser as its first-ever Chief Revenue Officer. She faces a monumental task: generating enough cash to cover a $2.5 billion burn rate and $1.4 trillion in planned AI infrastructure spending. The AI wars just got a lot more serious.
Global shipping runs on phone calls and spreadsheets, not real-time data. A new platform called Moddule is trying to use AI to connect the dots, but the industry is deeply skeptical of tech promises after past flops like blockchain.
Google DeepMind is partnering with the UK government to open its first automated research lab focused on material discovery. The facility, set to open in 2026, will use robotics and AI to search for new superconductors and other advanced materials.
Oracle’s much-hyped AI infrastructure demand wasn’t enough to save its quarterly results. The company missed revenue expectations, sending its shares tumbling and taking other major AI players down with it in premarket trading.
Intel’s “ANV” Vulkan driver in Mesa has merged a new shader VMA allocator. This is foundational work needed to enable Vulkan ray-tracing capture and replay, a crucial tool for developers. The feature is confirmed working with an older merge request in the RenderDoc tool.
A new patch in the GNU C Library (glibc) enables 2MB Transparent Huge Pages (THP) by default for memory allocations on AArch64 systems. This change aims to reduce page faults and kernel overhead, and has shown a consistent 6.25% performance improvement on SPEC benchmarks.
Oracle posted a huge beat on future contract value, but its stock still dropped. The market’s AI euphoria is fading, replaced by hard questions about execution and competition. It’s a reality check.
