AI News

Tool-using large language model (LLM) agents translate model outputs into consequential actions, turning indirect prompt injection into an execution-security problem. Existing defenses constrain tool calls, trajectories, or information flows, but often leave action semantics, state dependencies, and evidence relations implicit, limiting generalization across heterogeneous tools and complicating safe updates. We present \ours{}, a dynamic ontology-guided governance framework. \ours{} grounds heterogeneous calls in a shared semantic representation and combines an executable ontology with episode-level task state to derive deterministic ready, waiting, and forbidden decisions. A waiting decision identifies an unmet but recoverable obligation, enabling bounded replanning without expanding the agent's authorization boundary. \ours{} evolves its ontology offline from verified training executions. An offline LLM proposes instance-independent action--state axioms, which are accepted only if they address ontology errors exposed by training failures, preserve logical consistency and protected entailments, and pass held-out validation before the ontology is frozen. We evaluate \ours{} on AgentDojo, ASB, and AgentDyn under benign and indirect prompt-injection settings. \ours{} achieves a 0\% attack success rate (ASR) across all three benchmarks while improving task utility under attack compared with existing baselines. These results show that explicit semantic governance can strengthen execution security without indiscriminate blocking.

Global venture funding in October totaled $59 billion, a 45% jump from a year earlier, per Crunchbase data, in large part due to a record-setting $15.9 billion funding round for China’s DeepSeek.

Venture capital continues to flow into technologies that make artificial intelligence cheaper, more useful, and better able to operate in the physical world. From specialized AI chips and warehouse robots to distributed computing networks and brain-computer interfaces, the latest startup funding announcements reveal investors backing businesses with significant engineering requirements and potentially valuable positions in

Traversal, an AI startup building autonomous engineers to diagnose and resolve software outages, has raised $120 million in fresh funding at a $1 billion valuation, joining the growing ranks of AI unicorns. The round was co-led by Kleiner Perkins, Redpoint Ventures, and Khosla Ventures, with participation from Sequoia Capital, Battery Ventures, and Amex Ventures. The

L’impresa agentica apre una nuova fase della collaborazione uomo e macchina. Gli agenti artificiali gestiscono attività complesse in autonomia

Entire MacOS running in your terminal. Contribute to govijr/terminal-mac development by creating an account on GitHub.

CrowdStrike introduces the AI Kill Switch in Falcon Guardian, giving defenders human control to terminate rogue AI agents at the root.

PDK 1.0 due in early 2027.

TOKYO -- DRAM semiconductor memory prices have risen to roughly double their level at the end of last year, as manufacturers of the mainstay chips pri

China has deployed a world-first 800-volt (V) DC power system that can slash energy conversion losses and support more powerful data servers.

Join the discussion on this paper page

A paltry supply despite plenty of demand

Cybersecurity-focused Ballistic Ventures is seeking to raise $360 million for a new flagship fund and another $130 million for a separate investment vehicle.

CrowdStrike said it will buy Fiddler AI, which makes AI agent observability software, and Dreadnode, which helps companies automate AI security testing.

Chi investe in robotica, semiconduttori e tecnologie quantistiche? Una selezione di fondi deep tech che finanziano le startup

Il 5 novembre l'evento WMF in Silicon Valley

Argotec raccoglie 50 milioni di euro da NUO Space per sviluppare satelliti, software e AI. Un investimento strategico nel Tech Made in Italy.

Programmatically create and manage notebooks and sources with the Gemini Notebook API.

AMD delivers record quarter as Data Center segment revenue grows 97% year-over-year, driven by record sales of AMD Instinct GPUs and AM

Kimi K3.1 Pro has arrived, and the weights are dropping tomorrow.

Tomorrow, hear @Kimi_Moonshot 's Feihu Tang explain the architecture and decisions behind it.

He joins Austin Silveria, Kevin Cui and Zain Hasan from Together AI for a deep dive into long-running agents and production inference…