101 results for “AGI”
Trail of Bits tasked GPT-5.6-Cyber with escaping a QEMU/KVM VM on Debian 12. Over ~12 hours it broke out three times—kernel bug, unmarked libslirp, and a self-found 0-day chain—showing off-the-shelf VMs no longer contain such agents.
Google is expanding its AI & Economy Research Program with Nobel laureate Philippe Aghion as an advisor and new research directors. The hires track how AI reshapes work, productivity, and global technology diffusion.
Z.ai opened GLM-5.3-FlashX as a faster API SKU of Flash — claimed ~200 tokens/s, about 2.5× Flash list price, same multimodal stack. Coding Plan still Flash-only for now.
Hacktron chained a Discourse HEIF RCE with an OpenAI SSO flaw to reach employee ChatGPT and Codex sessions in under 72 hours. OpenAI paid $6,500 for the identity-side finding.
On 2 September 2026 the PyTorch Foundation released 2.14: nccl2 and NVGEMM land in-tree, fault tolerance becomes a first-class c10d feature, and Apple Silicon gains native linear algebra plus Metal kernels that close an 8.5× single-token decode gap versus the prior MPS path.
On 2 September 2026 Google DeepMind released Gemini 3.8 Flash for coding and long-horizon agents at unchanged 3.7 Flash pricing through December 2026. A separate 3.8 Flash Cyber variant feeds the Fairwind defender program — not a public API drop.
Stanford researchers introduced Paper2Agent, which turns a paper’s text, code, and data into a validated MCP server a chat agent can query in plain language. On AlphaGenome it built 22 tools in about 45 minutes for roughly $14.
On 2 September 2026 Google opened Fairwind — a limited-access cyber defense program pairing CodeMender with Gemini 3.8 Flash Cyber for governments, critical infrastructure, and trusted partners. It parallels OpenAI's Daybreak and Astra rollout on the same day.
IBM and Confluent brought Granite Time Series models to Early Access on Confluent Cloud, letting teams run forecasting and anomaly detection on streaming data via Flink SQL — without a separate ML stack.
Tencent's Hunyuan team released Hy-MT2 on 21 May 2026 — a new translation model family with instruction-following templates, three sizes up to a 30B-A3B MoE variant, and open weights plus IFMTBench. This is the next generation after the April Hy-MT1.5 phone quantizations.
Keycloak CVE-2026-18963 lets unauthenticated attackers reset any user's password without the email verification link. The flaw carries a 9.1 CVSS score and affects every deployment using the default password reset flow.
NVIDIA's NIM 2.0.12 stack boosts Nemotron 3 Ultra throughput to 1,997 tokens per second on four B200 GPUs, a 2.5x gain at 50 TPS per user. Enterprises can replay this profile using AIPerf.