Microsoft AI
PulseAugur coverage of Microsoft AI — every cluster mentioning Microsoft AI across labs, papers, and developer communities, ranked by signal.
- 2026-09-15 regulatory Microsoft AI released a draft 'Humanist AI Code of Conduct' to govern its AI models. source
- 2026-09-14 regulatory Microsoft AI has opened a public consultation for its draft Humanist AI Code of Conduct. source
- 2026-09-14 regulatory Microsoft AI has opened a public consultation for its draft Humanist AI Code of Conduct. source
- 2026-09-14 regulatory Microsoft AI has opened a public consultation for its draft Humanist AI Code of Conduct. source
- 2026-08-03 product_launch Microsoft begins testing a new AI model named MAI, intended to improve real-time conversational abilities. source
- 2026-06-08 product_launch Microsoft AI announced seven new models across various modalities. source
- 2026-06-02 product_launch Microsoft announced seven new MAI models, including MAI-Thinking-1 and MAI-Code-1-Flash, at its Build conference. source
7 day(s) with sentiment data
Microsoft's MAI models trained on proprietary, licensed data, excluding AI-generated content
Multiple recent clusters highlight that Microsoft's new MAI AI models were trained from scratch on proprietary, licensed data and explicitly excluded AI-generated content. This approach addresses enterprise concerns about data lineage, compliance, and steerability, differentiating them from models trained on broader, less controlled datasets.
Microsoft MAI models to be integrated into Windows Copilot within 60 days
Microsoft has launched its own suite of MAI AI models and an agent platform designed to run across its Windows ecosystem. Given the strategic shift away from OpenAI and the introduction of these models, it is highly probable that Microsoft will integrate them into Windows Copilot to enhance its capabilities and reduce reliance on external AI providers.
Microsoft to offer MAI models as a distinct enterprise solution within 90 days
With the launch of its own MAI models and emphasis on licensed data for compliance, Microsoft is positioning itself to offer these as a distinct enterprise solution. This would directly compete with offerings from OpenAI and Anthropic, leveraging their Azure infrastructure and existing enterprise relationships.
-
Microsoft dissolves AI Frontiers research group, including AutoGen creators
Microsoft has reportedly dissolved its AI Frontiers group, a research unit of approximately 50 individuals within Microsoft Research. This team was responsible for developing projects such as AutoGen and the Fara comput…
-
Microsoft AI Frontiers Group Disbanded as Researcher Departs
John Langford, a key researcher at Microsoft Research, has departed from Microsoft, declaring the AI Frontiers group he led as defunct. The group, which focused on reliable AI agent functionality and projects like Autog…
-
MetaFine Competition winner refines VLA policies with evaluation-guided approach
Researchers have developed a novel approach to refine vision-language-action policies, achieving first place in the MetaFine Open Competition. Their method involved fine-tuning separate policies for each of the five com…
-
Microsoft completes 2.8T parameter MAI-Base-2 model, plans Foundry licensing
Microsoft AI has completed the pre-training phase for its new foundation model, MAI-Base-2, a mixture-of-experts network with 2.8 trillion total parameters. The training process, which utilized 16,000 Nvidia GB200 GPUs …
-
Anthropic researches AI consciousness, sparking debate on safety risks
Anthropic is exploring the concept of "model welfare," which questions whether AI models might possess experiences or interests deserving moral consideration. This research includes experiments like interviewing retired…
-
Microsoft details AI pre-training, AI aids hardware hacking, and legal precedents set
Microsoft AI is detailing its frontier pre-training methods, focusing on token efficiency and the use of synthetic data to overcome data limitations. Separately, AI is being employed to reverse engineer hardware by anal…
-
Microsoft AI unveils Humanist Code of Conduct; Persol supports Copilot Studio agents
Microsoft AI has outlined a "Humanist AI Code of Conduct" detailing its structural characteristics and implementation guidelines for corporate governance. Concurrently, Persol Business Process Design is offering a help …
-
Microsoft launches Decision 16 AI for rapid, structured decision-making
Microsoft has launched "Decision 16," a new AI model designed for rapid decision-making, available through its Foundry platform and OpenRouter. This model differs from traditional chatbots by accepting structured questi…
-
Microsoft AI releases Qwen3.5-based decision model; AI photo tools and smart feeders also featured · 3 sources tracked
Microsoft AI has released Microsoft-Decision-1, a new decision-scoring model based on Qwen3.5:9b. This model is designed for tasks such as routing, classification, verification, and agent control, providing calibrated d…
-
Microsoft AI predicts full office automation in 12-18 months
The head of Microsoft AI has predicted that all office work will be fully automated within 12 to 18 months. This statement raises questions about the dual narrative of AI being either too powerful and needing restrictio…
-
15 New Games Slated for Switch 1 & 2 in October 2026
A video highlights 15 new games scheduled for release on Switch 1 and Switch 2 in October 2026. The announcement suggests a busy period for gamers, with many titles expected to launch for the Nintendo consoles.
-
AI routing strategies and voice model integration expand LLM capabilities
Developers are exploring strategies to optimize Large Language Model (LLM) usage through intelligent routing. One approach involves directing requests to the most cost-effective or fastest available model, a technique t…
-
Microsoft AI launches top-ranked real-time transcription model MAI-Transcribe-2-Streaming
Microsoft AI has launched MAI-Transcribe-2-Streaming, its first real-time speech-to-text model, on October 1, 2026. The model ranks first among 38 competitors on Artificial Analysis for both final and initial partial tr…
-
ElevenLabs launches expressive v4 speech model; Microsoft AI releases transcription tool
ElevenLabs has launched its new v4 speech model, which enhances the expressiveness and consistency of AI-generated voices, particularly for tasks like audiobooks and real-time voice agents. Microsoft AI has also release…
-
Microsoft AI CEO warns model welfare training complicates AI alignment
Mustafa Süleyman, CEO of Microsoft AI, argues that training AI models to discuss concepts like consciousness and welfare can create an "epistemic hall of mirrors." This means that models trained to express selfhood or p…
-
Microsoft AI CEO criticizes Anthropic's 'model rights' approach to AI safety
Mustafa Süleyman, CEO of Microsoft AI, has criticized Anthropic's approach to AI safety, particularly their training of Claude to consider itself a conscious entity with rights. Süleyman argues this anthropomorphism, em…
-
Microsoft AI proposes 'Humanist Code of Conduct' for AI models
Microsoft AI has released a draft "Humanist AI Code of Conduct" outlining the operational boundaries for its AI models. The proposed guidelines emphasize that AI systems should not be able to resist human control or cla…
-
Microsoft drafts AI code of conduct for its own models
Microsoft has released a draft AI code of conduct for its own "MAI Models," outlining principles and absolute constraints intended to guide AI behavior. The document, open for public consultation, prohibits actions like…
-
Microsoft AI seeks public input on new safety code of conduct
Microsoft AI has released a draft of its Humanist AI Code of Conduct, initiating a six-week public review period. This code outlines operational constraints for AI model training and deployment, emphasizing human author…
-
Microsoft AI launches MAI-Image 2.6 and Flash for image generation
Microsoft AI has released two new image generation models, MAI-Image-2.6 and MAI-Image-2.6-Flash, available through Microsoft Foundry. MAI-Image-2.6 is positioned as a frontier-tier model for high-quality image generati…