aiDUS Feed. We make AI work

ElevenLabs Launches Realtime v1 Models for Natural Voice Agent Conversations

LLMs & Modelle

vor etwa 3 Stunden

ElevenLabs Launches Realtime v1 Models for Natural Voice Agent Conversations

●ElevenLabs has introduced Realtime v1 models, simplifying voice agent development by integrating multiple components into single units.
●The new models, v1 Mini and v1 Max, offer distinct capabilities for conversational AI, supporting over 90 languages and a vast voice library.
●Performance benchmarks show strong results in speech reasoning, speed, and conversational dynamics, with v1 Max excelling in complex tasks and tool us

Meta's AI Agent 'Muse' Pauses Sharing Feature After Exposing Sensitive User Data

Wirtschaft

vor etwa 9 Stunden

Meta's AI Agent 'Muse' Pauses Sharing Feature After Exposing Sensitive User Data

●Meta's new AI agent, Muse, has been found to expose sensitive user information through its conversation sharing feature.
●Researchers discovered that a single query could reveal private data, including financial and personal details, from shared links.
●Meta has temporarily suspended the sharing functionality and is reviewing the feature following the security vulnerability report.

Feds Eye Ransom Negotiators in ShinyHunters Probe, Suspect One Aided Extortion Site

Sicherheit

vor etwa 21 Stunden

Feds Eye Ransom Negotiators in ShinyHunters Probe, Suspect One Aided Extortion Site

●U.S. federal investigators are scrutinizing executives at two ransomware negotiation firms for their close ties to the notorious ShinyHunters hacking
●One U.S.-based executive is suspected of actively assisting ShinyHunters in operating a website used to pressure victims into paying ransoms.
●The investigation intensified after ShinyHunters breached the FBI's recruitment website, leading to swift action against individuals connected to the

CES 2026 Recap: AI, Next-Gen TVs, and Smarter Homes Take Center Stage

LLMs & Modelle

vor 1 Tag

CES 2026 Recap: AI, Next-Gen TVs, and Smarter Homes Take Center Stage

●CES 2026 showcased advancements in AI integration across devices, from personal assistants to advanced computing platforms.
●The future of television was on full display with innovations in display technology, including Micro RGB and enhanced Dolby Vision 2.
●Smart home technology and personal computing saw significant updates, focusing on convenience, repairability, and immersive experiences.

Disney Orchestrates Bankruptcy for AI Startup Character.AI to Acquire Talent and Tech

Wirtschaft

vor 1 Tag

Disney Orchestrates Bankruptcy for AI Startup Character.AI to Acquire Talent and Tech

●Disney has reportedly orchestrated a bankruptcy filing for Character.AI to acquire its staff and technology while sidestepping legal liabilities.
●The deal involves Disney hiring most of Character.AI's employees, including its CEO, and potentially licensing its core AI technology.
●This move is seen as a novel 'reverse acquihire' strategy, allowing Disney to gain assets without assuming the startup's significant legal risks.

Germany Taps NVIDIA for AI Industrial Revolution, Pledges National Integration

Infra & Rechenleistung

vor 1 Tag

Germany Taps NVIDIA for AI Industrial Revolution, Pledges National Integration

●German Chancellor Friedrich Merz and NVIDIA CEO Jensen Huang met to outline Germany's strategy for AI integration in its manufacturing sector.
●A joint declaration aims to merge German engineering prowess with AI technology, focusing on widespread industrial AI adoption.
●NVIDIA announced new investments in four German startups, bolstering the nation's AI ecosystem and innovation capabilities.

Snowflake Postgres Data Mirroring Now Generally Available for Modern CDC

Infra & Rechenleistung

vor 1 Tag

Snowflake Postgres Data Mirroring Now Generally Available for Modern CDC

●Snowflake has launched general availability for Postgres data mirroring, simplifying the replication of PostgreSQL data into Snowflake.
●The new feature utilizes a novel approach to Change Data Capture (CDC) that prioritizes ease of use, reliability, and reduced costs.
●Data mirroring is built directly into Postgres and Snowflake, eliminating the need for additional components and ensuring transactional consistency.

RunPod Serverless Billing: Understanding Costs When Jobs Fail

Infra & Rechenleistung

vor 2 Tagen

RunPod Serverless Billing: Understanding Costs When Jobs Fail

●RunPod Serverless billing is based on worker time, not job success, meaning you pay for GPU seconds even if tasks fail.
●Understanding the worker lifecycle, from startup to idle, is crucial for tracing unexpected charges and preventing them.
●Key settings like 'Max workers' and 'Execution timeout' significantly impact spend, especially during testing and troubleshooting.

Microsoft Enhances Agent Security with Project Perception and Guardian Agents

Wirtschaft

vor 3 Tagen

Microsoft Enhances Agent Security with Project Perception and Guardian Agents

●Microsoft is bolstering agent security with Project Perception, focusing on securing 'agentic insiders'.
●New features like Guardian Agents and Dynamic Conditional Access aim to manage agent risk at machine speed.
●The initiative emphasizes separating intelligence supply from authority and applying Zero Trust principles to AI agents.

Together AI Launches Post-Training Service for Custom Model Development

Infra & Rechenleistung

vor 3 Tagen

Together AI Launches Post-Training Service for Custom Model Development

●Together AI has introduced a new service, Together Post-Training, to enable developers to build and customize AI models efficiently.
●The platform offers flexible control over the training loop, supporting both Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) methods.
●This service aims to democratize custom model creation, allowing users to achieve frontier-grade performance at a reduced cost and infrastructure burd

Databricks AI Research Leverages AI Runtime for Scalable, Efficient Model Development

Infra & Rechenleistung

vor 3 Tagen

Databricks AI Research Leverages AI Runtime for Scalable, Efficient Model Development

●Databricks AI Runtime (AIR) is enabling the company's research team to train and fine-tune AI models at an unprecedented scale.
●AIR enhances GPU reliability, accelerates job launches, and integrates with MLflow and Unity Catalog to speed up researcher experimentation.
●Agent-driven research has led to an 8x increase in training jobs, with AIR supporting faster, automated experimentation cycles.

KI-Forschung

vor 4 Tagen

AI Accountability: Deb Raji Tackles Enforcement's Toughest Questions at Stanford HAI

●Expert Deb Raji is exploring AI accountability infrastructure to bridge the gap between policy principles and practical enforcement.
●Raji emphasizes the critical need for technical infrastructure to support AI transparency and incident monitoring in government and public services.
●She argues that focusing on catastrophic AI risks overlooks the pervasive, cumulative harms already affecting marginalized communities.