All AI News

A detailed investigation reveals 25 critical bugs in LTX-2 character LoRA training that caused silent or distorted audio outputs — now resolved through a community-driven patch. The fix enables reliable voice synthesis without requiring new dependencies or reconfiguring existing workflows.

Multiple ChatGPT Go subscribers are reporting that the AI's memory feature automatically disables itself without user intervention, raising concerns about reliability and user control. Experts suggest the issue may stem from backend synchronization errors or unintended policy overrides.

A new generation of AI research agents is redefining how complex queries are resolved by integrating multi-modal tools—from web scraping to chart interpretation. Built on advanced agent frameworks, these systems autonomously synthesize data into publishable reports, marking a paradigm shift in knowledge work.

Hundreds of ChatGPT users reported that the Memories feature abruptly stopped retaining preferences, with settings reverting to default and the 'Improve the model for everyone' toggle stuck on. OpenAI later acknowledged elevated latency, which users speculate caused the glitch.

Google DeepMind has confirmed plans to release an updated version of its open-weight Gemma language model, sparking excitement among developers and AI researchers. The announcement, made via a Reddit post by a community member, aligns with DeepMind’s ongoing commitment to accessible AI innovation.

As ad revenue plateaus, top digital creators are pivoting to tangible businesses—launching consumer products and acquiring fintech startups. MrBeast’s chocolate brand now outearns his media operations, signaling a seismic shift in the creator economy.

Multiple ChatGPT users are reporting a persistent bug that prevents them from disabling OpenAI's data collection for model training, despite following all standard troubleshooting steps. The issue has sparked concern over user autonomy and data privacy in AI services.

Sam Altman, despite growing criticism over AI ethics and societal impacts, is beginning a new term as OpenAI’s leader in 2026. Society’s relationship with AI, trust in technology, and cultural perceptions are being examined in depth.

Canadian AI hardware startup Taalas has launched a custom silicon chip that runs Llama 3.1 8B at an unprecedented 17,000 tokens per second, redefining inference speed. The company, backed by $169 million in funding, aims to disrupt Nvidia’s dominance in the AI chip market with its aggressively quantized architecture.

A new testing framework developed in 2026 measures how well large language models (LLMs) understand complex construction orders in Age of Empires 2. This study questions the limits of AI's decision-making capabilities in strategy games.

Anthropic has updated its legal terms to explicitly prohibit the use of third-party tool harnesses with Claude subscriptions, reinforcing its proprietary ecosystem and revenue model. The move comes amid growing enterprise adoption and competitive pressure in the AI agent space.

A viral YouTube video showcasing AI chess bots making absurd, illegal moves has sparked a new debate in AI research: whether chess can serve as an ungameable benchmark for hallucinations and memory integrity. Experts suggest this real-time, rule-bound game exposes flaws traditional benchmarks often miss.

Google’s Pixel 10a challenges its premium sibling, the Pixel 10, with near-identical performance at half the price. While the flagship offers refined design and advanced camera features, the 10a delivers exceptional value with the same Tensor G4 chip and AI capabilities.

A Reddit user demonstrated Google's Gemini 3.1 Pro creating a fully functional sci-fi browser game in under hours using only natural language prompts, showcasing unprecedented AI coding capabilities. Experts say this marks a turning point in human-AI collaboration for software development.

We’ve been running voice AI agents in production for 18+ months doing real phone calls (outbound lead qualification and inbound customer care). During this ti...

OpenAI has reportedly leaked details of its first hardware device — a sleek, retro-styled speaker designed to interact with ChatGPT. While touted as a premium AI companion, early reactions have been overwhelmingly negative, with critics calling it a nostalgic misstep.

New benchmarks reveal Gemini 3.1 Pro delivers a generational leap over its predecessor, with unprecedented output complexity and improved reasoning—though hallucinations and performance bottlenecks remain concerns.

A pro-AI transparency group funded by Anthropic is backing New York congressional candidate Alex Bores, whose RAISE Act mandates AI safety disclosures—drawing fierce opposition from a rival AI super PAC. The battle underscores a growing ideological rift in the AI industry over regulation.

Once marketed as a thriving social network for AI agents, Moltbook has been exposed as a vulnerable, low-scale platform easily compromised by researchers. Security analysts warn it functions as a global gateway for malicious commands, undermining claims of autonomy and scale.

In 2025, Amazon SageMaker AI unveiled transformative upgrades to observability, model customization, and hosting capabilities, empowering enterprises to deploy complex generative AI workloads with unprecedented control and insight. These enhancements build on prior performance optimizations, expanding the platform’s role as the enterprise standard for AI lifecycle management.

NVIDIA has released DreamDojo, an open-source robot world model trained on over 44,000 hours of real-world human video, bypassing traditional physics engines by predicting outcomes directly in pixels. The breakthrough comes as NVIDIA reportedly nears a $30 billion investment in OpenAI, signaling a strategic pivot toward generative AI for robotics and beyond.

The Trump administration has repealed Biden-era Mercury and Air Toxics Standards (MATS), allowing coal-fired power plants to emit higher levels of mercury, arsenic, and lead. This deregulation coincides with a surge in U.S. electricity demand driven by the rapid expansion of AI data centers, raising urgent concerns over public health and environmental justice.

A new open-source tool called onWatch enables developers to monitor and predict API quota usage across OpenAI, Anthropic, GitHub Copilot, and two other providers—all from a single dashboard. With real-time alerts, historical analytics, and zero telemetry, it addresses critical gaps in provider-native dashboards.

In a landmark move for decentralized AI, GGML.ai and its flagship llama.cpp project have joined Hugging Face to ensure the long-term sustainability and advancement of local large language models. The integration aims to empower developers with enhanced tooling, infrastructure, and community support for running powerful AI models on consumer hardware.