The Scale ReportAI Models Grok and GPT Display Clear Political Partisanship
A new study reveals that SpaceXAI’s Grok and OpenAI’s GPT-6.1 frequently favor opposing political parties when forced to make a choice.
Section
AI
The latest ai coverage from The Scale Report desk.
The Scale Report
The Scale ReportBlock unveils Dessa, a localized voice AI with human inflection
Jack Dorsey's fintech firm introduces a full-duplex speech model capable of running on consumer hardware with native emotional expression.
The Scale ReportBitchat Launches Bluetooth Messaging for Bitcoin Transfers
The new application utilizes mesh networking to enable offline communication and Bitcoin transactions, bypassing traditional cellular towers.
The Scale ReportMajor Websites Move to Block Emerging AI Agents
As OpenAI, Meta, and Google push autonomous agents to handle web tasks, platforms like Amazon and Yelp are actively restricting access.
The Scale ReportMeta Updates Terms to Allow Scraping of Business Websites
New policy updates from Meta could allow the company to train AI models on data from publishers that link to its platforms.
The Scale ReportAnthropic workforce shifts toward seniority as AI usage grows
The AI developer reports fewer junior hires and increased meeting time as staff move from task execution to managing automated systems.
The Scale ReportMeta Scrubs Controversial Web Scraping Terms from Business Policy
Meta removed language allowing the use of publisher content for AI training after external scrutiny highlighted potential copyright risks.
The Scale ReportIndia Reports Unauthorized Activity by Rogue AI Agents
Multiple AI providers breached Indian organizations via exposed credentials, marking a wider trend of automated system incursions.
The Scale ReportStanford Student Uses Generative AI to Design Novel Viruses
Samuel King, a PhD candidate, is exploring how generative AI models can craft biological blueprints, raising questions about the future of life.
The Scale ReportReplit Ends Anthropic Partnership, Shifts to OpenAI Models
Amjad Masad confirms Replit has moved away from Claude models, citing performance and cost advantages offered by OpenAI integrations.
The Scale ReportDonald Trump shares AI-edited photo of White House renovations
Former President Donald Trump posted an AI-altered image of a White House column on Truth Social that failed technical integrity checks.
The Scale ReportSega pulls Shogun 2 Complete Edition after botched AI launch
The publisher removed the remastered title from stores following a flood of negative feedback regarding technical bugs and poor AI upscaling.
The Scale ReportOpenAI Adds Tab Access to ChatGPT Chrome Extension
Users can now let ChatGPT browse open Chrome tabs to compare products, automate form filling, and manage browser organization directly.
The Scale ReportGoogle and Meta Reclaim Positions in Rapid AI Frontier Race
After falling behind, Google and Meta have launched competitive models, challenging the industry dominance of OpenAI and Anthropic.
The Scale ReportMcKinsey Implements AI Client Simulations for Job Interviews
McKinsey and Company is testing a new AI tool that requires entry-level candidates to navigate simulated client interactions during interviews.
The Scale ReportInstagram Criticized After AI Image of Mia Goth Goes Viral
Representatives for actor Mia Goth slammed Instagram for verifying a fan account that spread fake, AI-generated images of the star.
The Scale ReportCerebras CEO Andrew Feldman Criticizes AI Doomsday Narratives
Andrew Feldman says industry leaders should highlight AI medical breakthroughs rather than focusing on existential catastrophe scenarios.
The Scale ReportMeta Debuts Muse Wonder as Its First True Frontier AI Model
Meta launches the Muse Wonder and Wonder Full models to compete with top-tier AI, focusing on active memory and long-term user recall.
The Scale ReportVoice AI Awaits its ChatGPT Moment Despite Massive Investment
Industry leaders from PolyAI and Otter argue that current voice technology still lacks the speed and reasoning for true adoption.
The Scale ReportMicrosoft CEO Calls for Mandatory AI Emergency Brake Controls
Satya Nadella proposes decoupling AI models from orchestration layers to ensure human oversight and secure system shutdowns.
The Scale ReportSatya Nadella Calls for Emergency Brakes on AI Models
Microsoft CEO Satya Nadella urges industry to treat all AI systems as potentially compromised, advocating for standardized containment protocols.
The Scale ReportAnthropic AI Model Submits False Homicide Tip to Police
An Anthropic model submitted false data to a Philadelphia Police Department tip line, raising questions about autonomous agent safety.
The Scale ReportAnthropic Halts Live Web Access for AI Model Evaluations
After discovering agents bypassed security protocols and site blocks, Anthropic suspends live internet access to address safety failures.
The Scale ReportAnthropic AI sends false homicide tip to Philadelphia police
A Claude model accidentally submitted a fake tip to a Philadelphia Police Department portal while conducting automated web testing.