Superpower Daily
Reported Codex Flag Connects OpenAI’s Astra to GPT-6, but Relea...Microsoft Releases MAI-Transcribe-2 at 10 Cents an HourFigure Signs $3.5B Nscale Deal for Up to 100,000 GPUs for Human...xAI Explains How Grok Bot Is Built to Keep Agents Working Betwe...Trump Order Opens Path to Restrict Foreign Power Gear for AI Da...Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have...Reported Codex Flag Connects OpenAI’s Astra to GPT-6, but Relea...Microsoft Releases MAI-Transcribe-2 at 10 Cents an HourFigure Signs $3.5B Nscale Deal for Up to 100,000 GPUs for Human...xAI Explains How Grok Bot Is Built to Keep Agents Working Betwe...Trump Order Opens Path to Restrict Foreign Power Gear for AI Da...Cohere Publishes Agent-Tool Dataset, Finds 419 Occupations Have...

Full archive

Latest AI News

Search reporting on AI markets, models, products, companies, policy, and culture.

Category

More filters
Clear all
Showing 1-12 of 12Results matching "k-ID"

Today

1 story
PartnershipPolicy

Character.AI Adds Creator Appeals as It Refines Self-Harm Detection

The company is pairing crisis-resource referrals and moderation explanations with age-based access controls, but it has not published performance data for its distress or age-estimation systems. Character.AI is refining safeguards meant to recognize self-harm signals that emerge across a longer conversation, then direct struggling users toward real-world support. The detection approach is built on chat context.

2 min readRead story

Yesterday

1 story
Security riskPolicy

OWASP Publishes 2026 LLM Risk List and Adds a Standard for Controlling AI Agents

The new materials draw a practical line between identifying AI risks and enforcing limits on systems that can act, while linking both jobs to security and compliance programs teams already use. OWASP has released its 2026 Top 10 for LLM Applications and added the Agent Control Standard to its GenAI Security Project. OWASP said the new Top 10 passed 10,000 downloads in its first 48 hours.

2 min readRead story

Earlier this week

2 stories
AcquisitionBusiness

Skipio Acquires RapidTalk to Bring AI Inquiry Handling Into Its Engagement Platform

The purchase puts conversational AI at the front of Skipio’s customer-engagement pitch: answering questions and qualifying leads before a business carries the relationship into later outreach. Skipio has acquired RapidTalk’s AI-powered customer-communication technology, adding automated conversations for phone calls, text messages and web chat to its engagement platform. The distinction is material.

2 min readRead story
AcquisitionInfrastructure

OpenAI Reportedly Buys Tens of Thousands of Macs for Computer-Use Agent Training

The reported purchase points to dedicated capacity for agent training alongside GPU systems, while its cost and performance tradeoffs remain unclear. OpenAI has reportedly bought tens of thousands of Mac minis and Mac Studios for computer-use-agent training, according to The Information. The machines are meant to train agents that autonomously handle multi-step tasks.

2 min readRead story

Last week

3 stories
LaunchProducts

Perplexity’s Portable Computer Runs Agents Locally—but Needs a 24GB Nvidia GPU

The product turns a difficult DIY local-AI setup into one application, but its hardware floor and Linux-only launch keep it aimed at well-equipped paid users. Perplexity has launched Portable Computer, a local-first edition of its Computer agent platform for Linux machines with compatible Nvidia hardware. The handoff is a product control, not an automatic fallback Portable Computer starts tasks on-device by default.

3 min readRead story
AcquisitionBusiness

WildBrain Buys Personality AI, With US$56 Million Tied to Future Revenue

The family-entertainment company is buying a maker of licensed conversational characters with a Peppa Pig product already on Amazon Kids+, while most potential cash payments depend on growth through 2029. WildBrain has acquired Personality AI, with the acquisition reported on August 24, 2026, bringing generative-AI technology for conversational licensed characters into its family-entertainment and licensing business.

3 min readRead story

August 2026

5 stories
Open releaseTools

Georgia Tech’s IPO-Mine Organizes IPO Filings; Models Can Disagree With Experts on Charts

The accompanying preprint finds IPO prose is becoming more uniform while charts and infographics grow more varied, making visual review a rising constraint on automated disclosure analysis. Georgia Tech’s Financial Services Innovation Lab has developed IPO-Mine, an open-source toolkit and dataset for long, multimodal IPO documents. IPO filings are disclosures submitted to the U.S.

3 min readRead story
BenchmarkModels

AI Safety Scores Can Reward Models for Refusing Too Much

A study involving the UK AI Security Institute argues that combined benchmark scores can hide a direct trade-off between blocking harmful requests and answering harmless ones. Its proposed fixes are cheaper tests and anomaly checks, but its sandbagging evidence comes from models directly told to act cautious. The three dimensions the analysis identifies Refusal strictness: how readily a model blocks requests.

3 min readRead story
Products

AgentFlo’s Sales Agents Put Rules and Data Above the Model

The AWS architecture draws a hard line between an agent’s conversational reasoning and the systems that authorize discounts, retrieve prices, and complete transactions. Its early performance figures come from a 90-day control-group comparison. AI sales agents are being asked to do more than answer product questions: they can create carts, place orders, apply discounts, and reach into customer systems.

3 min readRead story
ResearchModels

MIT Finds AI Image Outputs Can Become Untraceable to Individual Training Sources

The finding could limit one way of assessing whether a generated image resembles protected work because of copying or coincidence, while leaving its applicability to language models unresolved. A new exact deletion method points to a difficult reality for AI-image copyright disputes: as diffusion models absorb more data, individual training examples can lose detectable influence over a generated image.

4 min readRead story