Is superintelligence dangerous? Separate the two questions.

Here is what serious sources actually say, without the film-trailer language.

• A short, famous statement. In 2023 the Center for AI Safety published a single sentence signed by many leading researchers and company leaders: "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war."
• The International AI Safety Report 2026, chaired by Yoshua Bengio, defines loss of control as scenarios "where AI systems operate outside of anyone's control, with no clear path to regaining control". It adds: "Current systems lack the capabilities to pose such risks, but they are improving." It also describes an "evidence dilemma": AI systems are rapidly becoming more capable, but "evidence on their risks is slow to emerge and difficult to assess".
• OpenAI wrote in November 2025: "we treat the risks of superintelligent systems as potentially catastrophic" and "no one should deploy superintelligent systems without being able to robustly align and control them".
• Google DeepMind's 2025 safety paper groups risks into four areas: misuse, misalignment, mistakes and structural risks. It defines misalignment simply: the AI system "pursues a goal that is different from human intentions".
• Anthropic's CEO Dario Amodei set out five groups of risk in January 2026: autonomy (AI acting on its own goals), misuse for destruction, misuse for seizing power, economic disruption and indirect effects.
• Microsoft AI's Mustafa Suleyman wrote that such systems must be contained and aligned "not just once, but constantly, in perpetuity".

The common thread is not a prediction of doom. It is a claim that if systems far more capable than people emerge, controlling them would be hard, and the cost of getting it wrong would be very high.

A fair guide includes the people who think this is overblown. Their arguments, attributed:

• The technology may not get there by scaling. In a survey reported by the AAAI presidential panel in 2025, 76% of 475 respondents said scaling up current approaches is "unlikely" or "very unlikely" to yield AGI.
• Predicted take-off has not happened. MIT Technology Review reported that David Sacks, then the White House AI and crypto adviser, said "the doomer narratives were wrong", and that Sriram Krishnan, a senior White House AI policy adviser, called the notion of imminent AGI "a distraction and harmful and now effectively proven wrong". These are the views of those officials, and the administration's position has since moved toward promoting "Super Intelligence" as a label. See our explainer.
• Progress may have plateaued. Brookings analyst Mark MacCarthy argued in 2026 that "AI progress has plateaued rather than accelerated toward a breakthrough", which would weaken the case for strict controls.

The honest reading: experts disagree about the facts, not just the values. That is a reason for humility, and for making plans that work across several futures. See what the forecasts really say.

While the large debate runs, your business faces smaller, concrete AI risks that you can do something about this quarter.

The voluntary accord announced at the White House in September 2026 is, according to Nextgov, built on four layers: internal controls, an internal oversight team, independent outside auditors and an independent committee. You are not a frontier lab, but the structure scales down well.

• Know where AI is usedList every AI tool and feature in use, including free tools staff chose themselves.
• Name an ownerOne person is accountable for AI use, with the authority to say no.
• Limit access by defaultGive agents and tools only the permissions a task needs, and keep production data out of experiments.
• Keep a human in important decisionsMoney, hiring, safety and customer promises need a person to approve.
• Test before you trustTry tools on your own real cases and keep a record of failures.
• Get an outside viewAn occasional independent review catches what an internal team stops seeing.
• Review twice a yearThe tools and the rules are both moving fast.