Microsoft Copilot uses data as Cohere Transcribe leads Hugging Face

Key Takeaways

• GitHub Copilot, owned by Microsoft, will use user interaction data for AI training starting April 24th, requiring free and paid users to opt out via privacy settings.
• The Trump administration aims to pass the U.S. government's first comprehensive federal AI bill this year, led by Michael Kratsios and former AI czar David Sacks.
• AI security risks are growing beyond traditional firewalls, with new threats like prompt injections, data poisoning, and accidental data leakage requiring dedicated strategies.
• Sierra, co-founded by former Salesforce CEO Bret Taylor, launched Ghostwriter, an AI agent builder for tasks like customer service using text prompts.
• Cohere AI released Cohere Transcribe, an enterprise ASR model supporting 14 languages, which ranks first on the Hugging Face Open ASR Leaderboard with a 5.42% Word Error Rate.
• The AI industry faces potential economic risks from geopolitical instability and significant debt accumulation for data center expansion, raising concerns about a market crash.
• Leading AI labs, including OpenAI and Anthropic, acknowledge existential risks from AI, with probabilities of human extinction cited between 2% and 25%.
• AI-generated music is prompting debates about authenticity and calls for clear labeling or bans on streaming platforms.
• Companies' owned media, such as websites and blogs, are becoming crucial for AI visibility, with about 60% of LLM-cited material originating from corporate sources.
• David Sacks has concluded his role as Trump's crypto and AI czar but continues to advise on broader technology topics as co-chair of the President's Council of Advisors on Science and Technology (PCAST).

GitHub Copilot will use your data for AI training unless you opt out

Starting April 24th, GitHub Copilot will use interaction data from free and paid users to train its AI models. This includes inputs, outputs, and code snippets. Users of Copilot Business and Enterprise accounts are not affected. Those who previously opted out will keep their preference. All other users must manually change their privacy settings to prevent their data from being used for training. Many developers are reportedly unhappy with this policy change.

GitHub Copilot users must opt out to prevent AI training data use

GitHub, owned by Microsoft, will use user interactions with Copilot to improve its AI models, a common practice in the AI sector. The data includes inputs, outputs, code snippets, and suggestions. Users who have previously opted out of data collection for product improvements will not have their data used. All other users need to disable this feature in their privacy settings to prevent their data from being used for training. Microsoft states this will lead to more accurate and secure suggestions.

GitHub Copilot will use your code for AI training unless you opt out

GitHub announced that interactions with its Copilot tool will be used to train and improve its AI models. This applies to free and paid accounts, excluding Copilot Business and Enterprise. Users can opt out through their privacy settings. GitHub states this practice aligns with industry standards and will enhance model performance, leading to more accurate and secure code suggestions. Existing opt-out preferences will be maintained.

David Sacks ends role as Trump's crypto and AI czar

David Sacks has concluded his role as a special government employee serving as President Trump's crypto and AI czar. He will continue to serve as co-chair of the President's Council of Advisors on Science and Technology (PCAST). In this new capacity, Sacks plans to advise on a broader range of technology topics. He has been an influential figure in the Trump administration since the president's second term began. Sacks is also a founding partner at Craft Ventures.

White House aims for first comprehensive federal AI bill

The Trump administration is working to pass the U.S. government's first comprehensive AI bill. The proposed law aims to protect children using AI, prevent data center rate hikes, and preempt state regulations. Science adviser Michael Kratsios and AI and crypto czar David Sacks are leading the effort. They hope to get a bill signed into law this year. The White House released a blueprint for the bill last week, which has received mixed reactions from industry leaders and lawmakers.

AI security risks grow beyond traditional firewalls

Artificial intelligence and large language models (LLMs) introduce new security challenges that traditional methods cannot handle. Prompt injections, where attackers trick AI into overriding its purpose, and indirect injections hidden in external data are significant threats. Data poisoning, feeding malicious data to train AI, can corrupt its behavior. AI models can also accidentally leak private information from their training data. Companies need dedicated AI security strategies to address these novel vulnerabilities.

Sierra unveils AI agent builder Ghostwriter

AI startup Sierra, co-founded by former Salesforce CEO Bret Taylor, has launched Ghostwriter. This new product allows users to build AI agents for tasks like customer service using simple text prompts. The self-service nature of Ghostwriter is a key feature for Sierra, potentially reducing operational costs. This development aims to make AI agent creation more accessible and efficient for businesses.

AI boom faces risks from global instability and debt

The global economy's reliance on AI is threatened by geopolitical instability and massive debt. The AI supply chain, from chips to data centers, depends on materials from the Middle East, now destabilized by war. Disruptions in the Strait of Hormuz impact oil, gas, and crucial chip manufacturing components. Major AI companies are accumulating significant debt to fund massive data center expansions. This precarious situation, with interlocking fragilities, could lead to a market crash.

Prepare for AI economic bubble crash now

Policymakers must prepare for a potential AI economic bubble crash, learning from past crises like 2008. The AI industry faces a mismatch between trillions invested and current revenues, funded by public money through various financial instruments. This creates systemic risks, similar to the housing bubble, where interconnectedness and complex financial engineering obscure dangers. A crash could severely impact the global economy, affecting retirement accounts and investments. Proactive planning is crucial to mitigate the fallout.

Cohere AI launches advanced speech recognition model

Cohere AI has released Cohere Transcribe, a new Automatic Speech Recognition (ASR) model for enterprise use. The model uses a hybrid Conformer encoder and Transformer decoder architecture for high accuracy. It supports 14 languages and ranks first on the Hugging Face Open ASR Leaderboard with a 5.42% Word Error Rate. Cohere Transcribe outperforms models like Whisper Large v3 and handles long-form audio by processing it in 35-second segments.

AI's existential risks and economic promises raise concerns

Leading AI labs like OpenAI and Anthropic acknowledge significant risks, including potential human extinction from AI technology, with probabilities cited between 2% and 25%. This raises questions about the ethics of developing such powerful technology. Simultaneously, AI is driving massive investment and economic growth, but concerns exist about a potential bubble. The dual narrative of existential threat and economic disruption presents a challenging outlook for AI's future.

AI music raises questions about art and authenticity

The rise of AI-generated music, like the song 'I Run' by HAVEN. and the band The Velvet Sundown, is blurring the lines between human artistry and machine creation. While some AI music gains significant traction on platforms like TikTok and Spotify, critics argue it lacks soul and devalues human artists' skills. There are calls to ban AI music from streaming platforms or at least label it clearly. The debate questions whether AI music is genuine art or merely 'slop.'

Owned media is key for AI visibility

In the age of AI, companies' own websites and blogs are becoming crucial for visibility in large language model (LLM) outputs. Approximately 60% of material cited by LLMs comes from corporate-owned sources. Communications teams need to structure their websites to be optimized for AI. This strategy helps brands control their narrative and appear in AI-generated answers, shifting focus from borrowed audiences to owned expertise. Freshness and content structure are key factors for AI optimization.

Sources

• GitHub Reverses Course And Will Train AI On Your Copilot Data Unless You Opt Out
• Users will share data for AI training with GitHub Copilot, unless they opt out
• GitHub’s Copilot will use you as AI training data, but you can opt out
• David Sacks says his time as Trump's crypto and AI czar has ended
• The elusive AI bill that the White House wants to land
• Beyond Basic Security: How AI is Rewriting the Safety Playbook
• AI Startup Sierra Unveils Self-Service Agent-Building Product
• The AI Boom Wasn’t Built for the Polycrisis
• We Must Prepare For an AI Bubble Now
• Cohere AI Releases Cohere Transcribe: A SOTA Automatic Speech Recognition (ASR) Model Powering Enterprise Speech Intelligence
• AI has the worst sales pitch I've ever seen
• AI in the Music Industry: Art or Slop?
• Owned in an AI world