Other AI

A team from MemTensor (Shanghai), HONOR Device and Tongji University released a preprint describing MemPrivacy, an edge-cloud framework that pseudonymizes sensitive values on-device using typed placeholders so cloud

A jury in Oakland rejected Elon Musk's 2024 lawsuit against OpenAI and CEO Sam Altman after two hours of deliberation on May 18, 2026.

A May 2026 Mandiant survey finds attackers using automation and AI to speed intrusion workflows, forcing defenders to operate at machine pace.

Wave Terminal, now in beta for Linux, macOS and Windows, bundles terminal, file manager, process viewer, web view and an AI assistant into a single open-source app to centralize developer workflows and reduce tool

A how‑to guide by Jessica Lau (May 18, 2026) catalogs seven AI security risks and offers concrete mitigations for teams embedding generative AI in browsers, inboxes, and other work tools.

Contributing writer Jack Wallen tested three AI tools on May 18, 2026 to generate a Hyprland configuration; only OpenAI’s Codex produced a remotely usable hyprland.conf for Hyprland 0.55.

The Vatican finalized an encyclical on artificial intelligence titled "Magnifica Humanitas," signed by Pope Leo XIV on May 15, 2026; he will present it personally on May 25.

Security platforms and open-source projects are being overwhelmed by a surge of low-quality, AI-generated bug reports that have forced program suspensions and prompted new validation and triage measures.

VCs are increasingly rejecting AI startups based on nine founder red flags, favoring teams with durable product moats and trustworthy leadership over polished demos.

Anthropic will brief finance ministries and central banks on thousands of severe security flaws its Claude Mythos Preview model identified, after a request from Andrew Bailey;

Researchers will present new evidence at the IEEE Symposium on Security and Privacy showing that imperceptible audio manipulations can hijack large audio — language models (LALMs).

Engineers built a production — grade Model Context Protocol (MCP) server on AWS so LLM clients can query a B2B intelligence dataset of more than one million company profiles while minimizing risk to production systems.

On May 18, 2026, Cloudflare and Stripe launched a protocol via Stripe Projects (open beta) that automates account creation, identity, and payments so coding agents can wire accounts, purchase domains, and deploy apps

A writer with no coding experience used a Claude Pro subscription to prototype a communal database that logs everyday administrative 'sludge'—small bureaucratic hassles that accumulate into larger burdens.

A May 18, 2026 analysis finds 34 AI startups generating almost $80 billion in annual revenue — up 112% in six months — with Anthropic overtaking OpenAI largely on demand for its AI coding tools and the two firms together

NVIDIA published a 4‑bit pretraining methodology centered on NVFP4, a microscaling FP4 format supported natively on Blackwell Tensor Cores, and validated it by training a 12‑billion‑parameter hybrid Mamba‑Transformer

At a May 12 Cannes press conference where Demi Moore was serving on the festival jury, the actress urged artists to “work with” AI rather than fight it and argued the technology can’t replicate the “soul” of true art,

Rivington Labs Inc., operating as AirOps, introduced Quill on May 13, 2026: an autonomous AI agent that monitors, updates and drafts brand content so companies remain visible in answers returned by generative AI search

OpenAI quietly bought audio‑AI startup Weights.gg earlier this year, folded the roughly six‑person team into internal product groups, and took the startup’s site offline on April 1, 2026;

A May 15, 2026 hands‑on guide publishes runnable code and step‑by‑step notes for assembling an MCP‑style routed agent that discovers tools, selectively exposes capabilities, plans tool use, executes calls, and injects

On May 17, 2026 Oppo's Multi‑X team published source code and a technical report for X — OmniClaw, an open‑source Android agent that runs on physical phones and aligns vision, screen contents, voice and text

Kathy Oneto recommends a six-question checklist employees should use before accepting new work goals.

At its Code with Claude developer conference in San Francisco, Anthropic introduced “dreaming,” a research‑preview feature for Claude Managed Agents that periodically reviews sessions and memory stores to surface

A hands — on tutorial uses llmcompressor to apply post-training quantization to an instruction — tuned model (starting from FP16) and compares FP8 dynamic, GPTQ W4A16, and SmoothQuant combined with GPTQ W8A8, reporting