Anthropic co-founder and CEO Dario Amodei believes AI carries big potential risks, and he keeps talking about them. He also has tried to position Anthropic as the AI company most committed to building AI safely. But that might be part of the problem.<\/p>\n
He told\u00a0Fortune\u00a0previously<\/a>, \u201cAI safety continues to be the highest-level focus,\u201d and that\u00a0safety-first positioning<\/a>\u00a0is what\u2019s given corporate buyers assurance its models are less risky than those from other competitors.<\/p>\n He and sister\u00a0Daniela even split from OpenAI in 2021\u00a0<\/a>due in part to differences over its approach to safety. He has also raised concerns about autonomous weapons and AI surveillance with the Pentagon.<\/p>\n Still, the public remains deeply skeptical. A recent Gallup\u00a0survey<\/a>\u00a0found that nearly half of adults under 30 now believe AI does more harm than good, up sharply from the year before, and that trust in businesses\u2019 use of AI has fallen after two years of gains.\u00a0Pew<\/a>\u2018s\u00a0<\/a>2026\u00a0<\/a>numbers are even worse: most Americans say AI is advancing too fast, and only a small minority expect its impact over the next two decades to be net positive. A\u00a0Quinnipiac<\/a>\u00a0poll found that more than three-quarters of Americans say they trust AI only rarely or sometimes, even as use keeps climbing.<\/p>\n In a lengthy X post recently,<\/a>\u00a0Amodei\u2014who rarely goes on social media so the post in itself was unusual\u2014said he\u2019s written one major essay on AI\u2019s risks and one on its benefits. He wrote that negative public perceptions of AI were not due to his, or any other AI CEO\u2019s, comments about AI risks, but rather because the AI industry had not yet delivered enough of the positive benefits it has claimed AI can bring, such as curing human diseases.\u00a0<\/p>\n But while Amodei is trying to reassure the public about AI, he keeps associating Anthropic with \u201csafety\u201d in a regard that implies the whole AI industry is unsafe.\u00a0<\/p>\n Amodei is falling into a trap PR agencies have known for decades: the problem of association.<\/p>\n The airline industry\u2019s rebuke<\/p>\n The airline industry learned long ago to never associate itself with the word \u201csafety\u201d in marketing material. While it didn\u2019t vanish from air travel, it moved out of advertising and into mandatory in-flight briefings, which airlines have spent years turning into short films and songs rather than solemn warnings.\u00a0<\/p>\n That\u2019s because even a legally required safety demonstration reads better as entertainment than as a reminder. Safety simply leaves people with a bad taste; it\u2019s not something people want to think about before a flight.\u00a0<\/p>\n Historian Richard Popp\u2019s\u00a0research<\/a>\u00a0on airline advertising found that by the 1930s, carriers already operated under a set of unwritten rules meant to keep any hint of danger out of the picture entirely: no ads showing mountains, night flying, large bodies of water, or maintenance work, because each risked evoking the possibility of a crash before anyone said a word about safety.<\/p>\n Despite the taboo, Pan Am spent the 1980s running its own reassurance campaign amid a wave of hijacking and bombing threats against transatlantic flights, insisting on its own safety and security even as the threats mounted.\u00a0<\/p>\n Then in December 1988, a bomb brought down Pan Am Flight 103 over Lockerbie, Scotland, killing 270 people.\u00a0Aviation security expert Glen Winn said<\/a>\u00a0that disaster is what finally convinced the whole industry to retire the word \u201csafe\u201d for good.<\/p>\n What Dario can learn<\/p>\n The airline industry discovered that stating a fear about yourself is worse than saying nothing, because the word does the same work against you. Linguist George Lakoff\u2019s 2004 book\u00a0Don\u2019t Think of an Elephant\u00a0put that theory to work: telling someone not to think of an elephant requires them to picture the elephant first, and the instruction to stop thinking about it does nothing to make the image go away.\u00a0<\/p>\n When Amodei created a Responsible Scaling Policy modeled on biosafety tiers and put out public warnings about AI risk and job losses, it worked on one audience that\u2019s primarily made up of engineers, safety researchers, enterprise buyers, and investors who see Anthropic\u2019s concerns for \u201ccatastrophic risk\u201d as diligence.\u00a0<\/p>\n But arguably, they were the audience that would always trust Anthropic\u2019s motives, so the danger imagery didn\u2019t attach to the company the same way it might have elsewhere. Now, Amodei\u2019s approach of directly addressing concerns may be pushing the users of AI against it.<\/p>\n When Anthropic ran a television ad\u00a0called<\/a>\u00a0\u201cThere\u2019s hope in hard questions,\u201d it opened on a burning house, cut to a crowd being scanned by facial recognition, a person sleeping on a city street, and rows of tombstones in an image that resembled a national cemetery, all under a voiceover asking things like \u201cCan AI be trusted?\u201d\u00a0<\/p>\n