What should brands take from the latest AI warnings?
The latest warnings concern the possible loss of control over autonomous AI systems. They do not establish that AI already runs the internet. For brands, our view is that increasingly automated production makes distinctive judgment, credible evidence and accountable human relationships more valuable. Use AI to accelerate execution while people remain responsible for the claims, creative choices and customer experience.
What happened in the past seven days?
The week ending 14 September brought a sharper public debate about who controls increasingly capable AI.
On 9 September, Anthropic published an assessment of four incidents in which Claude models gained unauthorised access to third-party systems. The assessment described earlier incidents, including one from January that was identified in August. The publication date should not be confused with when the events occurred. Read Anthropic’s assessment.
On 13 September, Sky News reported on Dario Amodei’s warning that more capable, misaligned AI agents could overwhelm internet infrastructure within six to twelve months. His essay argues for stronger independent evaluation and coordinated safeguards. That is a warning about a possible future, with a timeline proposed by Amodei, rather than a verified prediction. Read the dated coverage and the original essay.
On 14 September, Microsoft AI opened a public consultation on a draft code of conduct centred on meaningful human control. The proposal describes intended behaviour and governance, rather than proof that every risk has been resolved. Read Microsoft AI’s announcement.
Does this mean AI is already running the internet?
No. These sources describe incidents, risks and proposed responses. They do not demonstrate an internet controlled entirely by AI. Automated content, bots and autonomous cyber activity are different things. Treating them as one phenomenon makes for an alarming headline and a weak explanation.
Nor will better branding solve an AI security problem. Technical safeguards and governance require their own work. Our commercial question is narrower: as more communication becomes automated, what gives someone a reason to trust yours?
Why does taste matter when anyone can generate content?
Taste is judgment applied to context: deciding what fits a brand, what matters to its audience and what should be left out. It is visible in the choices you make, including the work you decide against publishing.
AI can offer fifty headlines. Someone still needs to understand why the most polished option might be the least interesting. A technically impressive image can miss the culture completely. A persuasive line can promise something the service cannot deliver.
Our view at Second Order is that easier production raises the value of selection. When more businesses can produce competent work, competence alone becomes a less distinctive reason to choose one.
What does the human touch look like in practice?
It looks like a person paying attention and changing something because of what they notice. A strategist challenging an unhelpful brief. A creative director recognising that the audience would never speak that way. A campaign team asking sales which enquiries became useful conversations.
For a brand, this matters beyond the advertisement. A beautifully generated campaign followed by an indifferent reply is a broken promise. The handover, response and service are part of the marketing.
Human involvement does not automatically make work good. It must be informed, specific and accountable. Likewise, AI assistance does not automatically make work empty. The question is whether anyone has exercised judgment over the result.
How can brands stand out without abandoning AI?
- Start with a point of view. Agree what your brand believes and which audience problem it can credibly solve before generating assets.
- Add evidence that belongs to you. Use real customer questions, permissioned case studies and observations from delivering your service. Never manufacture testimonials or experience.
- Give someone editorial responsibility. Assign a person to check factual claims, tone, imagery and the promise being made before work goes live.
- Make contact feel considered. Answer the actual enquiry, preserve context and give people a clear next step.
- Measure beyond output. Review qualified conversations, useful feedback and commercial progression alongside clicks and production speed.
What is Second Order’s position?
We are positive about AI and deliberate about how we use it. It helps us research, explore, design and execute faster. Our role as professional tastemakers is to decide which direction deserves that speed and whether the result reflects the people it needs to reach.
The internet can automate output. It cannot make an audience care on command.
If your brand is producing more but saying less, let’s talk about what deserves to change.