Strategic Foresight Group Unveils AI Report, Calls For Global Rules On Frontier AI
· Free Press Journal
Mumbai/Geneva: A new report by the Strategic Foresight Group (SFG) has called for global cooperation to manage the most dangerous risks posed by advanced artificial intelligence (AI), despite growing geopolitical competition among major powers.
Visit betsport.cv for more information.
Titled "The Essential Convergence: Global Compact on Extreme AI Risks," the report was launched in Geneva on July 6. The event was attended by leading AI experts and heads of international organisations, including the International Federation of Red Cross, Inter-Parliamentary Union, International Campaign to Abolish Nuclear Weapons (ICAN), ICT4Peace Foundation and the Geneva Centre for Security Policy.
Growing Global Agreement
The report says countries may disagree on technology, trade and AI leadership, but they are increasingly finding common ground on preventing the most serious AI threats.
India Plans National Cybersecurity Strategy For Finance Sector With AI Risk FocusThe United States and China continue to compete over semiconductors, supply chains and AI leadership. Europe has introduced strict AI regulations, while India and many Global South countries focus on fair access to advanced AI systems.
Despite these differences, policymakers across the world are beginning to agree on the need to address extreme AI risks.
Four Major AI Risks
According to the report, governments broadly agree on four key risks that require urgent attention.
The first is cybersecurity, where advanced AI systems could be used to attack critical infrastructure or sensitive defence systems.
The second is the misuse of AI in developing biological or chemical weapons, making it easier for terrorists or rogue groups to create dangerous pathogens.
The third concern is large-scale misinformation and manipulation, where AI could influence public opinion on an unprecedented scale.
The fourth risk involves highly autonomous AI systems behaving in unexpected ways that humans may not be able to control.
‘AI Must Serve Humanity, Not Rule It’: Sanjay Pradhan Pushes For Ethical Leadership At Mumbai EventCommon Standards Emerging
The report notes that several countries are already introducing similar approaches to identify the most powerful AI systems.
South Korea, China and lawmakers in the United States are using computing power, or compute thresholds, as one way to define frontier AI models requiring greater oversight.
The report suggests that AI systems trained at around 10²⁶ to 10²⁷ FLOPs could serve as an international reference point for applying additional safety measures.
Testing Before Release
One of the report's strongest recommendations is that advanced AI models should undergo safety testing before being released to the public.
The United States recently introduced a voluntary framework encouraging AI developers to provide frontier models to the government for review before launch.
China, the UAE and the European Union have also introduced different systems requiring safety checks and risk assessments for highly capable AI models before deployment.
Reporting Safety Incidents
The report also recommends mandatory reporting of serious AI safety incidents within about 72 hours.
Examples include theft of AI model weights, harmful cyberattacks carried out by AI, failures of biological safeguards and situations where AI systems behave beyond their intended limits.
The report says similar reporting rules already exist in industries such as aviation, nuclear safety and public health.
‘AI Tsunami’ Is Coming; World Dangerously Underprepared: Anthropic CEO Dario AmodeiFocus on Cooperation
SFG President Sundeep Waslekar, the principal author of the report, said these proposals are not aimed at creating a global AI regulator. Instead, they focus only on preventing the most catastrophic AI risks through practical cooperation.
The report concludes that while countries will continue competing in AI development, they share a common interest in preventing disasters that could threaten everyone. It argues that cooperation on these limited but critical issues can become the foundation for safer AI worldwide.