Meredith Whittaker, president of the Signal Foundation, argues that the immediate danger from AI lies in careless use, including in intelligence and military settings, rather than machines turning against people. She also says Signal would leave any country that forced it to weaken its encryption, putting a clear line in the sand as governments push for more access to private messages.
Watch Desk analysis
What happened
In a BBC interview, Whittaker said fears of AI turning on humanity were not anchored in science. She pointed instead to risks from systems used carelessly, including an alleged case in which a chatbot invented information about a Chinese vessel and reportedly contributed to preparations for a US military response. That account was reported by CNN, according to the BBC, and Whittaker described it as a near-miss.
Whittaker also connected her concerns to the use of AI in military targeting. She said the US missile attack on a school in Iran relied on Palantir’s Maven system and outdated intelligence. The US military has not released its investigation findings; Palantir told the BBC there was no evidence its software was at fault and said it was not responsible for the underlying data or intelligence deficiencies.
On encryption, Whittaker told the BBC that Signal would leave a market rather than expose users to surveillance if a government required it to weaken the service. She cited the risks to people who rely on private communications, including those facing abuse or state surveillance.
Why it matters
Whittaker’s argument is a useful counterweight to the more cinematic AI-risk debate: the danger she describes is not a machine acquiring its own ambitions, but people relying on unreliable systems in decisions with serious consequences. Her examples also raise questions about where responsibility lies when AI-generated information enters intelligence work.
The encryption position brings a related policy trade-off into view. Governments argue that access to messages can help protect children, while Signal says weakening encryption would expose users more broadly. The dispute is not an abstract tussle between privacy and safety; it is about who bears the risk when private communications become easier to access.
Our read
Whittaker is making a forceful case for taking deployment choices more seriously than science-fiction scenarios. Her claims about specific military incidents remain claims, not settled findings, and the distinction matters. But her broader point deserves attention: “human in the loop” is not much comfort if the information reaching that person is wrong, or if the system’s role is difficult to scrutinise.
For readers, the useful question is less whether AI will rebel than where it is already being trusted, what information it supplies, and who can challenge its output. That is a less theatrical debate, which is often where the consequential one is hiding.
What to watch
- What the US military’s investigation says about the Minab school strike and the intelligence behind it.
- Whether governments pursue rules requiring encrypted services to provide access to messages.
- How military and intelligence users assess AI-generated information before acting on it.
Discussion spark: When AI-generated information enters high-stakes intelligence work, should responsibility sit mainly with the people who act on it, the organisations deploying the systems, or the companies building them?
Sources and evidence
- 'Careless use of AI is the real threat – not the ghost stories' – BBC (9 October 2026, 07:05 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.