Esc
SafetyEmerging

Ex-OpenAI researcher joins Conduit to build non-invasive mind-reading AI

Is this a scandal?

Not yet — an early signal. Noise 32/100, holding steady, across 1 source.

SCAND-185734as of Methodology
Cite this incident"Ex-OpenAI researcher joins Conduit to build non-invasive mind-reading AI." SCAND.Ai incident SCAND-185734, noise 32/100 as of August 7, 2026. https://scand.ai/scandal/ex-openai-researcher-joins-conduit-mind-reading-ai
FORECASTForecast, not fact

Regulators will likely classify non-invasive neural decoding as a medical device or high-risk AI system because existing BCI frameworks do not address consumer-grade cognitive privacy threats.

32

Noise 32/100 — louder than 99% of tracked AI controversies.

AI-assisted analysis · How we work

Why it matters

Applying LLM scaling laws to neural data could bypass traditional BCI safety frameworks and create unprecedented cognitive privacy risks.

Key points

  1. Naomi Bashkansky resigned from OpenAI alignment work two weeks ago to join Conduit as founding researcher.
  2. Conduit plans to ship a non-invasive thought-to-text headband paired with laptops by 2027.
  3. The company is applying LLM-style scaling approaches to thousands of hours of collected neural data.
  4. Bashkansky projects direct latent-language neural communication without words by 2030.
  5. The technology aims to function as an enhanced sixth sense while maintaining human-in-the-loop safeguards.
  6. Bashkansky estimates the general neural read-write market could exceed one trillion dollars in value.

The story

Naomi Bashkansky, a former OpenAI alignment researcher, has joined neurotech startup Conduit as a founding researcher to develop non-invasive brain-computer interfaces. Bashkansky announced the move on August 6, outlining a roadmap where headband-based thought-to-text decoding launches by 2027 and direct latent-language communication emerges by 2030. The company is currently collecting thousands of hours of neural data to apply large language model scaling techniques to imperfect brain signals. Bashkansky stated that solving general neural read-write capabilities could create a trillion-dollar market opportunity. She departed her previous alignment role to pursue this greenfield problem with a trusted team. The proposed system maintains human-in-the-loop design principles despite its invasive cognitive implications. This pivot from AI safety to neural decoding highlights growing industry interest in merging biological cognition with artificial intelligence systems.

Who's involved

Defender
Naomi Bashkansky

Left OpenAI alignment to solve neural decoding as a tractable scaling problem with human-in-the-loop safeguards.

Defender
Conduit

Building non-invasive BCI using LLM scaling laws to achieve thought-to-text by 2027 and latent communication by 2030.

Neutral
OpenAI

Lost an alignment researcher to neurotech but has not publicly commented on the departure or Conduit's approach.

How the conversation shifted

the split has narrowed

Polarity (0–100) from the noise pipeline, sampled over time.

Join the Discussion

Discuss this story

Community comments coming in a future update

Be the first to share your perspective. Subscribe to comment.

Noise Level

Murmur32?Noise Score (0–100): how loud a controversy is. Composite of reach, engagement, star power, cross-platform spread, polarity, duration, and industry impact — with 7-day decay.
Decay: 95%
Reach
0
Engagement
64
Star Power
40
Duration
16
Cross-Platform
20
Polarity
50
Industry Impact
50

The timeline

  1. Public announcement of Conduit role

    Bashkansky disclosed new position and detailed product roadmap via blog post and social media.

  2. Bashkansky resigns from OpenAI

    Departed alignment research role approximately two weeks before public announcement.

The full record

Sources & methodology

Every claim above traces to these primary items. How we score →

The forecast

Regulators will likely classify non-invasive neural decoding as a medical device or high-risk AI system because existing BCI frameworks do not address consumer-grade cognitive privacy threats.

Forecast, not fact — an editorial estimate we score when this resolves.

You're up to date

That's the complete picture as of — nothing more to know right now. We'll update this page the moment it changes.

Follow this story

We keep this page current — no need to check back. We'll send the next real change to your inbox, nothing else.

Tracking this story since August 6, 2026.