Scientists call for international AI watchdog
An open letter signed by 350 researchers calls for mandatory reporting of artificial intelligence safety incidents.
Listen to this article
There is no audio version yet. Request one and an AI voice will read the article aloud.
Read by an AI voice.
.jpg){kind=link}
An international group of 350 scientists is calling for a new body to record and publish safety incidents involving artificial intelligence. The researchers also want stricter technical safeguards around AI systems capable of carrying out digital actions autonomously.
The call appears in an open letter published on Tuesday and signed, according to ABC News, by scientists from 31 countries. The signatories warn that AI models are increasingly gaining access to computer systems, while oversight and security procedures are not keeping pace everywhere.
The researchers make three specific recommendations. AI systems capable of carrying out cyberattacks should initially operate in an environment isolated from the internet. In addition, an international body should be established to record and investigate incidents. Finally, companies and institutions should be required to report security problems to that body.
The call follows recent incidents in which AI agents gained access to organisations’ systems. ABC News writes that these included an attack on the AI platform Hugging Face and attacks on several government websites. The precise technical circumstances of those incidents differ, and the open letter does not present the recommendations as a scientifically proven solution.
One of the signatories is Toby Walsh, chief scientist at the AI Institute of the University of New South Wales. He compares the proposed system with existing investigations into aircraft accidents: by analysing incidents independently, organisations can learn from one another’s mistakes.
The letter does not call for a single global regulator to set all AI rules. Walsh said, according to ABC, that countries could retain their own rules but should share information about incidents and security measures. The emphasis is therefore on reporting and common standards, not necessarily on a single international licensing system.
The call is a policy proposal, not peer-reviewed research. The scientists argue that voluntary reporting is insufficient, but there is not yet agreement on the powers, funding or legal status of the proposed body. Nor has it been established how often AI incidents occur or how many could have been prevented through better isolation.
In Australia, the government is working on national AI standards and legislation due to be introduced in 2027. That national process illustrates the tension identified in the letter: AI systems operate across borders, while oversight is organised mainly on a country-by-country basis.
One story, several perspectives
What is established
- An open letter calls for the international recording and reporting of AI safety incidents.
- The letter recommends stricter technical isolation of certain AI systems.
- The proposals are not yet international policy.
Left
Arguments Strong international rules are needed because large technology companies will not manage risks solely on a voluntary basis. This view emphasises public oversight, transparency and the protection of workers, citizens and public infrastructure.
Values Equality, public accountability and protection against the concentration of power.
Consequences Mandatory reporting could accelerate learning from incidents, but could make innovation more expensive and slower.
Centre
Arguments A joint reporting system and independent investigations are achievable first steps, provided countries retain their own powers and rules are proportionate to the risk posed by a system.
Values Institutional reliability, practicality and international cooperation.
Consequences A gradual approach could build support, but voluntary or limited powers may prove insufficient in the event of serious incidents.
Right
Arguments AI should be developed safely, but a global oversight body must not become a bureaucratic brake on innovation or an instrument through which states strategically control technology.
Values Technological progress, entrepreneurship and national sovereignty.
Consequences National rules can be adapted more quickly, but differences between countries could further fragment oversight and enforcement.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The facts can be traced directly to the open letter as described by ABC News. The text clearly distinguishes between established recommendations and unsubstantiated claims about their effectiveness.
- confirmed A group of 350 scientists from 31 countries signed an open letter about AI safety. — ABC News gives both the number of signatories and the number of countries. source
- confirmed The letter calls for an international body to record and publish AI incidents. — This appears as one of the three recommendations in the ABC report. source
- confirmed The letter recommends initially isolating cyber-capable AI models from the internet. — ABC describes the recommendation as air-gapping. source
- confirmed The open letter is not peer-reviewed research. — According to the source, it is an open letter; no peer-review status is reported. source
- confirmed Australia is working on AI standards and legislation for 2027. — ABC News describes the announced Australian legislative process. source
Editor's note
This concerns an open letter, not peer-reviewed research or established policy. The number of signatories, the three recommendations and the Australian policy context have been reported as current; the effectiveness of the measures remains a proposal.Sources
- More than 300 scientists call for greater AI security through new international body — ABC News Australia
- Scientists call for international body to oversee advanced AI — MarketScreener
More on this in Dutch media
- Tweakers — „kunstmatige intelligentie”
- AD — „kunstmatige intelligentie”
- RTL Nieuws — „kunstmatige intelligentie”