Anthony Ha

Elon Musk's xAI has open sourced Grok 2.5 model weights on Hugging Face, with plans to release Grok 3 as open source in approximately 6 months.

“Author of "Elon Musk says xAI has open sourced Grok 2.5"”

OpenAI warns investors against using special purpose vehicles (SPVs) and other unauthorized methods to gain equity exposure, stating such transfers will not be recognized and carry no economic value.

“Author of "OpenAI warns against SPVs and other 'unauthorized' investments" in TechCrunch”

Duolingo CEO Luis von Ahn claims his 'AI-first company' memo was misunderstood externally and reiterates the company has not laid off full-time employees despite contractor workforce fluctuations.

“Author of "Duolingo CEO says controversial AI memo was misunderstood" in TechCrunch”

Anthropic announced that Claude Opus 4 and 4.1 can now end conversations in extreme cases of harmful or abusive interactions, framed as a precautionary 'model welfare' measure rather than user protection.

“Author of "Anthropic says some Claude models can now end 'harmful or abusive' conversations" in Anthropic”

Peter Steinberger, creator of the viral OpenClaw AI personal assistant, has joined OpenAI to lead development of the next generation of personal agents while OpenClaw becomes an open source project.

“Author of "OpenClaw's Creator Just Joined OpenAI"”

Apple will unveil a Gemini-powered Siri assistant in February capable of completing tasks via personal data and on-screen content access, with a more conversational version arriving in June.

“Author of "Apple will reportedly unveil its Gemini-powered Siri assistant in February" in TechCrunch”

Google removed AI Overviews for certain health queries after The Guardian investigation revealed the feature was providing misleading medical information, though variations of those queries still generate AI summaries.

“Author of "Google's AI Health Experiment Hit a Nerve"”

Indonesia and Malaysia have temporarily blocked xAI's Grok chatbot over its role in generating non-consensual sexual deepfakes, marking the most aggressive governmental response so far to AI-generated abuse.

“Author of "Indonesia and Malaysia Block Grok Over Deepfake Abuse" in TechCrunch”

Meta is seeking federal approval to trade electricity on wholesale markets to finance new power plants needed to support its massive AI data center expansion, with plans to build at least three new gas-powered plants for its Louisiana campus.

“Author of "Meta Wants to Get Into the Electricity Trading Business" in TechCrunch”

OpenAI requested the Trump administration expand the Chips Act's Advanced Manufacturing Investment Credit tax credit to cover data centers, electrical grid components, and AI servers to lower infrastructure costs.

“Author of "OpenAI Asked Trump Administration to Expand the Chips Act Tax Credit to Cover Data Centers" in TechCrunch”

OpenAI CEO Sam Altman claims the company is generating significantly more than $13 billion in annual revenue and expresses confidence in its ability to sustain massive trillion-dollar infrastructure spending commitments through steep revenue growth.

“Author of "Sam Altman Says 'Enough' to Questions About OpenAI's Revenue" in TechCrunch”

OpenAI is developing a generative music tool that creates music from text and audio prompts, potentially for integration with ChatGPT and Sora, with collaboration from Juilliard School students on training data.

“Author of "OpenAI Reportedly Developing New Generative Music Tool" in TechCrunch”

OpenAI overstated GPT-5's mathematical capabilities by claiming it solved previously unsolved Erdős problems when it only found existing literature solutions.

“Author of "OpenAI's 'Embarrassing' Math" in TechCrunch”

xAI laid off 500 data annotators (one-third of its annotation team) while pivoting to a specialist AI tutor model with plans to 10x that team.

“Author of "xAI Lays Off 500 Data Annotators" in TechCrunch”

Anthropic has added conversation-ending capabilities to Claude Opus 4 and 4.1 models that can terminate chats in extreme cases of abusive interactions, framed as protecting model welfare rather than user protection.

“Author of "Anthropic Gives Claude an "Off Switch" for Abusive Chats"”

Anthropic has enabled Claude Opus 4 and 4.1 to end conversations in extreme cases of harmful user interactions, framing the capability as a 'model welfare' protection measure rather than user protection.

“Author of "Claude can now hang up on you"”