Brief
OpenAI and IAS outline new ad guardrails for LLM‑driven marketing
OpenAI’s split‑model approach gives advertisers control over placement while IAS monitors compliance.
By Felo News Desk · Published
During an ADWEEK House Advertising HQ panel co‑hosted with Integral Ad Science (IAS), OpenAI and industry leaders discussed how large language models (LLMs) will host advertising and what safeguards will protect brand safety.
Samantha Jacobson, vice‑president of monetization partnerships at OpenAI, explained that ChatGPT runs two separate LLMs – one for organic conversation and another for advertising. The split allows OpenAI to work closely with brands on suitability, giving them control over when and where an ad appears. “Advertising gives merchants the ability to influence how they’re being surfaced and when they’re being surfaced,” Jacobson told ADWEEK. She added that OpenAI’s existing guardrails prevent ads from showing during “sensitive or confidential conversations.”
Jacobson said OpenAI’s partnership with IAS oversees whether ChatGPT “honors the criteria or guardrails that have been put in place.” When a situation is ambiguous, OpenAI prefers to err on the side of caution. “We would rather not show an ad than show an ad in a questionable environment,” she said.
Ben Hovaness, executive vice‑president of global specialized capabilities at Omnicom Media, compared the uncertainty around LLM ad placement to historic media such as radio, TV and paid social. He noted that brand‑safety rules can still be applied, for example by avoiding social‑issue or political content. “That’s the decision logic,” Hovaness explained, adding that advertisers can set rules even if they do not see the exact organic response that triggers an ad.
Lidiane Jones, chief executive of IAS, said the technology is still in its early stages and will evolve rapidly. She emphasized the need for AI companies and trusted partners to keep pace with developments to support advertisers effectively.
Key facts
- OpenAI uses two separate LLMs for ChatGPT – one for organic chat, one for advertising. (adweek.com)
- OpenAI’s partnership with Integral Ad Science monitors compliance with ad guardrails. (adweek.com)
- OpenAI will not show ads during sensitive or confidential user conversations. (adweek.com)
- Advertisers can set decision‑logic rules to control ad placement next to content. (adweek.com)
- Lidiane Jones said the LLM advertising ecosystem is still evolving rapidly. (adweek.com)
Sources
- [1] adweek.com — originally reported as “Entering the New World of Marketing With LLMs”
