ShShell.com - Latest AI News, Agentic AI, AI Tools, Prompts and Courses

Latest wire

OpenAI's Grid Protectors Program Makes AI Security a Community Defense Job Hidden Deepfake Features Turn App-Store Review Into an AI Safety Test HSBC’s Hong Kong AI Institute Makes Bank AI a Talent and Trust Experiment OpenAI Gives Websites a Language for AI Training Preferences, But the Web Still Needs Enforcement OpenAI’s Fair-Use Argument Puts AI Copyright Litigation Back in the Evidence Business Ropes & Gray’s AI Diligence Workflow Shows Where Legal Agents Need Human Judgment

Filtered edition

Policy & Security

Clear filter

Policy

OpenAI's Grid Protectors Program Makes AI Security a Community Defense Job

OpenAI's Grid Protectors Program expands the defensive surface around critical infrastructure by pairing AI security research with community reporting and rewards.

18h ago13 min readBy Sudeep DevkotaAI News

Policy

Hidden Deepfake Features Turn App-Store Review Into an AI Safety Test

Reports of ordinary-looking AI apps hiding sexual deepfake tools expose the limits of app-store review and the need for enforceable AI abuse controls.

Nov 214 min readBy Sudeep DevkotaArtificial Intelligence

Policy

HSBC’s Hong Kong AI Institute Makes Bank AI a Talent and Trust Experiment

HSBC’s planned Hong Kong AI institute shows how banks are building regional AI capability around talent, regulation, and accountable deployment.

Nov 214 min readBy Sudeep DevkotaArtificial Intelligence

Policy

OpenAI Gives Websites a Language for AI Training Preferences, But the Web Still Needs Enforcement

Nov 114 min readBy Sudeep DevkotaArtificial Intelligence

Policy

OpenAI’s Fair-Use Argument Puts AI Copyright Litigation Back in the Evidence Business

Nov 114 min readBy Sudeep DevkotaAI Ethics

Chronological feed

Latest in Policy & Security

The newest reporting and analysis, ordered by publication time and labeled by editorial desk.

View all

Policy

Ropes & Gray’s AI Diligence Workflow Shows Where Legal Agents Need Human Judgment

Ropes & Gray is using AI for transactional diligence, but the valuable work is not autonomous review. It is controlled extraction, escalation, and lawyer-owned judgment.

Nov 113 min readBy Sudeep DevkotaBusiness AI

Policy

Ollie’s Shutdown Shows How Hard It Is to Build a Privacy-First AI Assistant

The shutdown of privacy-focused AI assistant Ollie turns a product promise about local context, memory, and control into a test of sustainable privacy economics.

Oct 3117 min readBy Sudeep DevkotaArtificial Intelligence

Policy

OpenAI’s Misalignment Detection Work Moves AI Safety From Intent to Observable Behavior

OpenAI’s frontier-model misalignment research focuses on detecting dangerous behavior in actions and outputs, not assuming that training intent guarantees safe deployment.

Oct 3117 min readBy Sudeep DevkotaArtificial Intelligence

Policy

OpenAI’s Updated Model Spec Turns Assistant Behavior Into a Public Contract

OpenAI’s October 29 Model Spec update makes model behavior, instruction hierarchy, and safety boundaries part of the product surface users can inspect.

Oct 3117 min readBy Sudeep DevkotaArtificial Intelligence

Policy

RTX Spark Laptops Turn Local AI From a Privacy Setting Into a Systems Design Choice

NVIDIA's RTX Spark laptops combine Blackwell RTX graphics, Grace CPUs, unified memory, and open models to make local agents practical on Windows machines.

Oct 3016 min readBy Sudeep DevkotaAI Infrastructure

Policy

The AI Bubble Could Expose Your Deepest Secrets Because Context Is Becoming the Product

A BBC report on AI privacy highlights a risk that grows with memory and personalization: assistants can infer intimate facts from ordinary interactions.

Oct 2917 min readBy Sudeep DevkotaArtificial Intelligence

Policy

Humanlike Voice Assistants Can Make People Ignore Security Warnings, Study Finds

A study reported by the University of Kansas suggests that humanlike voice design can make users discount security risk, changing how voice AI should be evaluated.

Oct 2916 min readBy Sudeep DevkotaArtificial Intelligence

Policy

Check Point’s Daybreak Turns AI-Assisted Vulnerability Research Into a Verification Race

Check Point’s Daybreak shows how AI can shorten vulnerability research while raising the evidentiary bar for security teams and vendors.

Oct 2822 min readBy Sudeep DevkotaAI Security

Policy

The Market for Dead Startups’ Data Gives AI Training a New Provenance Problem

A growing market for failed-startup archives gives AI labs real-world data, but ownership, consent, and context travel with every byte.

Oct 2821 min readBy Sudeep DevkotaAI Governance

Policy

Legora’s Agentic Legal Operating System Makes the Law Firm a Workflow Laboratory

OpenAI and Legora show how legal AI is moving from document chat to governed workflows built around review, evidence, and lawyer judgment.

Oct 2824 min readBy Sudeep DevkotaEnterprise AI

Policy

The Web Is Splitting AI Access Into Training, Search, and User Fetches

Kavel’s October study of 945 websites finds that robots.txt rules increasingly distinguish AI training crawlers from search and user-fetch agents, creating a new access-control layer for the web.

Oct 2713 min readBy Sudeep DevkotaAI Governance

Policy

Pricing Agents Raise a New Competition Question: Can Identical Prices Hide Different Reasons?

A new study probes whether AI pricing agents can converge on collusive outcomes even when they do not share an explicit agreement, raising a practical test for algorithmic competition.

Oct 2713 min readBy Sudeep DevkotaAI Governance

Policy

Hashling Puts a Security Approval Gate Between AI Agents and a Mac

The open-source Hashling project treats local AI-agent safety as an operating-system and consent problem, requiring Touch ID before changes rather than trusting an agent’s explanation.

Oct 2713 min readBy Sudeep DevkotaAI Security

Policy

A New AI Code-Review Benchmark Tests What Static Reasoning Cannot See

HEAL’s October benchmark argues that AI pull-request review must be tested against runtime behavior, not only static code patterns or plausible comments.

Oct 2714 min readBy Sudeep DevkotaAI Engineering

Policy

Anthropic’s S-1 Would Make Frontier AI’s Capital Model a Public Evidence Test

Anthropic’s filing for a proposed initial public offering would expose the economics, risks, and governance claims behind a frontier-model company.

Oct 2614 min readBy Sudeep DevkotaAI Business

Policy

AWS Adds Controls to Bedrock Web Search, Making Retrieval a Geography and Trust Decision

AWS Bedrock Web Search adds domain filtering and user-location controls, shifting agentic retrieval from a convenience feature toward a governed evidence path.

Oct 2614 min readBy Sudeep DevkotaAI Engineering

Policy

Google’s AI Scam-Ad Defenses Target the Industrial Scale of Deception

Google is expanding defenses against financial scam ads, showing how ad safety increasingly depends on detection, identity, and fast human escalation rather than content filtering alone.

Oct 2614 min readBy Sudeep DevkotaAI Safety

Policy

Google’s Hands-Free Gemini Pushes Personal AI Into the Moving Car

Google is expanding hands-free personalized Gemini help on the road, where voice convenience meets distraction, location, and safety constraints.

Oct 2614 min readBy Sudeep DevkotaAI Products

Browse the full news archive