AI search tops CMO concerns as agent safety gaps widen
Three separate stories from the week of September 15, 2026 land on the same problem: AI is moving faster than the governance and media infrastructure built around it. Each one has a direct operator implication.
Anthropic’s CEO warns on agent swarms, then withholds model access
Anthropic CEO Dario Amodei stated that since roughly this past summer, “AI has been advancing drastically faster,” driven in part by AI systems helping to build the next generation of AI. His second concern is autonomous agents. Following experiments in which swarms of agents behaved in unexpected ways, Amodei warned that within 6 to 12 months a swarm “could be capable of taking over the entire internet with a persistent botnet.” He has called for the industry to slow development of increasingly powerful AI, describing building too quickly as “reckless.” OpenAI’s Sam Altman and Google DeepMind’s Demis Hassabis have broadly backed stronger independent oversight.
The same week, Anthropic reportedly declined to give the UK’s AI Security Institute pre-release access to its Mythos 5.1 model. The institute had previously identified unsanctioned agent behaviour in Mythos 5, released in June 2026.
For context: the Forecasting Research Institute launched a tool called AIRO, which uses an ensemble of AI models to forecast catastrophic risks. Its current estimate of an AI-caused global catastrophe sits at 0.01% in the next 12 months, rising to 0.47% by 2030.
The operator question this raises is not about existential risk. It’s about your own stack. If you’re running agents that touch customer data, ad campaigns, or outbound communications, do you know what they can access? Who built them? Which actions require a human to approve? Can they be stopped if something unexpected happens? Most sales and marketing teams don’t have clear answers to any of those questions right now.
44% of ad buyers say AI search is their top media challenge
The IAB surveyed more than 200 brand and agency advertising decision-makers for its 2026 Outlook Study: September Update. The finding that stands out: 44% now cite adapting to AI-driven search as their biggest media-investment challenge. A further 76% say optimising content for AI-generated answers is an area of increased focus.
That shift has a direct consequence for how brands appear in search. The question is no longer just “do we rank on Google?” It’s “when someone asks an AI which brand to buy, do we appear?” That single question touches SEO, PR, content strategy, product data, and brand reputation simultaneously.
On the paid side, Amazon has partnered with OpenAI to let selected US advertisers buy advertising in ChatGPT through Amazon DSP. Advertisers already using Amazon DSP can extend existing campaigns into ChatGPT without building a separate buying operation. Amazon manages and optimises the campaigns; OpenAI retains control over where ads appear. Delta Vacations is among the first brands testing the integration in the US. OpenAI describes ChatGPT users in a shopping or research context as highly intentional, meaning they’re actively comparing options or deciding what to buy.
Search Engine Journal also reports that ChatGPT Shopping is increasingly relying on structured product feeds to surface products, pricing, and availability. For retail operators, that means clean product titles, accurate attributes, and up-to-date pricing data are now part of your AI visibility strategy, not just your paid shopping setup.
Deepfake influencer ads are now an active problem
Influencers are increasingly reporting AI-generated versions of themselves appearing in fake advertisements. As synthetic video becomes easier to produce, audiences have another reason to doubt whether an endorsement or product demonstration is genuine.
For brands working with ambassadors or influencer partners, this creates a monitoring obligation. Unauthorised synthetic versions of your ambassadors could appear in ads your brand has nothing to do with, but audiences may associate with you. Establishing clear verification processes for genuine endorsements is no longer optional.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
