BotBeat
...
← Back

> ▌

IntelIntel
FUNDING & BUSINESSIntel2026-08-05

OpenAI Researcher Launches Conduit Intelligence to Build Brain-to-Text AI

Key Takeaways

  • ▸OpenAI founding researcher Naomi Bashkansky leaves to launch Conduit Intelligence
  • ▸Conduit focuses on thought-to-text AI models for brain-computer interfaces
  • ▸Company's mission explicitly centers on safe, beneficial AGI development
Source:
Hacker Newshttps://naomibashkansky.com/↗

Summary

Naomi Bashkansky, a founding researcher at OpenAI, has announced the launch of Conduit Intelligence, a new startup focused on developing thought-to-text models that enable direct brain-to-computer interfaces. The company aims to create AI systems capable of translating human neural activity directly into written text, representing a novel frontier at the intersection of neurotechnology and generative AI.

Bashkansky's transition from OpenAI reflects a strategic pivot toward brain-computer interface (BCI) technology and its potential to transform human-AI interaction. She explicitly frames the venture's mission around ensuring that Artificial General Intelligence development benefits humanity—a clear focus on AI safety and alignment alongside technical innovation. The move underscores a growing trend of senior AI researchers and executives departing major tech companies to launch specialized startups pursuing novel AI capabilities and modalities.

  • Exemplifies talent migration from large AI labs to specialized startups pursuing frontier capabilities
Generative AIMachine LearningStartups & FundingAI Safety & Alignment

More from Intel

IntelIntel
PARTNERSHIP

Safe Superintelligence Inc. and NVIDIA Forge Long-Term Strategic Partnership to Accelerate Safe AGI Development

2026-07-31
IntelIntel
PARTNERSHIP

Intel Licenses Atom CPU Technology to Startup RosaicLabs in Unusual RTL Deal

2026-07-31
IntelIntel
RESEARCH

Prime Intellect Unifies 365,000 Agentic AI Tasks Under Single Evaluation Framework

2026-07-29

Comments

Suggested

Independent / Open SourceIndependent / Open Source
RESEARCH

Interlock: A Runtime Firewall That Assumes Prompt Injection Already Won

2026-08-05
AnthropicAnthropic
RESEARCH

New Security Benchmark Reveals Dramatic Variations in AI Model Safeguards

2026-08-05
Hugging FaceHugging Face
OPEN SOURCE

Developer Releases Empty 16.5T Parameter Model to Satirize AI's Scaling Obsession

2026-08-05
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us