CNN Just Ran One of the Worst AI Puff Pieces We’ve Ever Seen in Our Lives About That “AI Actor” Tilly Northwood18m ago How Credit Genie keeps codebase docs fresh with OpenWiki20m ago Closing The Gap Between AI Ambition And Data Reality21m ago ElevenLabs signs its first major-label deal, licensing Universal music for a fan remix platform27m ago Top Pentagon official dismisses viral AI warning from an Anthropic researcher27m ago Arlequin AI raises €28M to build novel AI models that learn complex relationships at scale30m ago SpaceX sues to block release of tax-break records for its Texas Terafab project31m ago Salesforce closes Fin and eyes $2bn Listen Labs, Business Insider reports32m ago Google brings Gemini for desktop app to Windows32m ago AI’s Hidden Apprenticeship Tax (And How Firms Can Avoid It)36m ago Abacus.AI Releases Three Open-Weight Smaug Models for Agentic Workloads41m ago Clay raises $115m at a $7.1bn valuation, more than double its 2025 price42m ago Before AI Agents Take Action, Enterprises Need A New Data Rule Book51m ago Feature Engineering in Scikit-Learn: A KDnuggets Cheat Sheet51m ago

CNN Just Ran One of the Worst AI Puff Pieces We’ve Ever Seen in Our Lives About That “AI Actor” Tilly Northwood18m ago How Credit Genie keeps codebase docs fresh with OpenWiki20m ago Closing The Gap Between AI Ambition And Data Reality21m ago ElevenLabs signs its first major-label deal, licensing Universal music for a fan remix platform27m ago Top Pentagon official dismisses viral AI warning from an Anthropic researcher27m ago Arlequin AI raises €28M to build novel AI models that learn complex relationships at scale30m ago SpaceX sues to block release of tax-break records for its Texas Terafab project31m ago Salesforce closes Fin and eyes $2bn Listen Labs, Business Insider reports32m ago Google brings Gemini for desktop app to Windows32m ago AI’s Hidden Apprenticeship Tax (And How Firms Can Avoid It)36m ago Abacus.AI Releases Three Open-Weight Smaug Models for Agentic Workloads41m ago Clay raises $115m at a $7.1bn valuation, more than double its 2025 price42m ago Before AI Agents Take Action, Enterprises Need A New Data Rule Book51m ago Feature Engineering in Scikit-Learn: A KDnuggets Cheat Sheet51m ago

AI Research2 min reading time

Yet Another Safety Leader at OpenAI Has Left

GizmodoPublished Jul 11, 2026

Johannes Heidecke, OpenAI's Head of Safety Systems, has left the company amid a series of departures in safety leadership roles. OpenAI faces increased safety demands due to faster model training and release cycles, leading to organizational changes in safety teams.

More on this story

OpenAI’s Head of Safety Is Leaving the Company
Jul 11· 2 sources

#johannes-heidecke #model-training #openai #release-cycles #safety-systems

More in AI Research

AI Research4 min read

An Anthropic researcher just quit, saying OpenAI and Anthropic are 'gambling with our lives'

Covered by 12 sources1d ago

AI Research3 min read

Worried Anthropic researchers warn that AI ‘could kill all humans’

Covered by 8 sources1d ago

AI Research4 min read

Suno trained its v6 AI music models with help from Warner and BMG

Covered by 5 sources1d ago