Anthropic is giving METR badges, desks, laptops and risk-team-level access to independently test whether its safety promises and incident reporting hold up, as OpenAI signals it will follow suit.
Qwen3.6 now reaches 10.05 tokens per second on two 8 GB GPUs—up from 1.44—by keeping its active experts in CUDA VRAM and using a 9 GB/s-plus-3 GB/s dual-SSD tier that reads expert weights roughly 33% faster while preserving bit-identical output.
OmniVoice now generates 24 kHz speech across 600-plus languages from 3–10-second voice samples, though cross-lingual clones retain the accent of the reference speaker’s language.
Mach Industries raises a $600 million Series C extension after paying $50 million in cash and equity for Exquadrum, outbidding at least eight suitors and expanding into solid rocket motors through its new Mach Energetics unit.
Apple says its upcoming Live Rewind feature captures the previous 15 seconds after a Digital Crown double-press, then uses the S11 chip’s Secure Enclave to process and immediately delete the audio, while Siri Recap offers toggleable high-level conversation summaries without transcripts, speaker labels, or timestamps.
Apple embeds AI across its entire product lineup, unveiling the S11 and A20 Pro chips alongside privacy-focused Audio Intelligence features like Siri Recap and Live Rewind, which deliver conversation summaries and on-demand transcripts without recording audio.
Qualcomm and Amazon ink a landmark $4 billion AI chip deal, with AWS gaining warrants to purchase 25 million Qualcomm shares as part of a multi-generational partnership targeting Nvidia's dominance in the data center market, potentially unlocking up to $60 billion in chip purchases.
Humanoid robots are boxing, dancing, and playing soccer at IFA 2026, with models now available for purchase up to $99,999 as AI rapidly closes the gap between science fiction and reality.
AI visionary Ray Kurzweil joins neurotech startup Subsense as an advisor, backing its groundbreaking noninvasive brain-computer interface that delivers charged nanoparticles to the brain via nasal spray, controlled by a magnetic headset — positioning the $27M-funded company as a consumer-friendly alternative to Elon Musk's Neuralink.
Extropic's new Z1T chip delivers over 100x energy efficiency gains compared to Nvidia's H100 GPU by using sparse, in-memory computing and a disaggregated inference pipeline, consuming just 294.52 nJ per token versus 40.9 µJ on the H100, while also achieving faster per-token latency — with open-source weights and training recipes …