OpenAI hires top marketing exec for ads business as Mythos/GPT-5.5 cyber capabilities draw Washington scrutiny
OpenAI's week was a mix of commercial maturation and political headwinds. On the commercial side, the company is actively recruiting a senior marketing executive specifically to promote its advertising business — the clearest signal yet that ads are no longer an experiment but a strategic revenue pillar OpenAI wants to brand-build around. The hire follows months of quieter expansion of ad placements inside ChatGPT and OpenAI's growing roster of media partnerships.
The political headwind is the Politico story on GPT-5.5's cyber capabilities. UK AISI's red-teaming found that GPT-5.5 could fully take over a corporate network in 3 of 10 attempts — exactly half the 6/10 rate AISI reported for Anthropic's Mythos. The number is lower than Anthropic's but still high enough that policymakers in Washington are reportedly alarmed. Source D notes the disclosure is being read alongside Anthropic's Mythos-rollout plans as evidence that frontier model capability has crossed a threshold regulators were not prepared for.
The talent-shuffle subplot is Andrej Karpathy's reported move to Anthropic's pre-training team, which prompted a viral AI-generated 'Michael Scott onboarding' parody on X. The parody is meme-fodder; the substance is that Anthropic continues to attract high-profile ex-OpenAI and ex-Tesla researchers in numbers that would have been unthinkable two years ago. For OpenAI, the talent-magnet narrative is shifting in a direction Sam Altman cannot fix with a tweet.
The ChatGPT product surface continues to expand. Source A flagged that Codex is now in preview inside the ChatGPT mobile app, letting users monitor a Codex session running on a connected Mac, plus 'Appshots' for app-window context and a generally available 'Goal mode' for outcome-defined tasks. OpenAI is also separately spending heavily on a deployment-services arm — Source B framed the $4B figure as a signal that enterprise AI now requires dedicated consulting infrastructure to operationalize. The aggregate picture: OpenAI is becoming a more conventional enterprise software company even as its core models trigger more conventional regulatory anxiety.