Everywhere Insiders 66: BRICS Divisions and AI Security Risks

Irina Tsukerman is a New York-registered human rights and national security lawyer, geopolitical analyst, and journalist. She earned a B.A. in International/Intercultural Studies and Middle East Studies from Fordham University in 2006 and a J.D. from Fordham University School of Law in 2009. She is president of Scarab Rising, Inc., editor-in-chief of The Washington Outsider, and fellow of the Jerusalem Center for Security and Foreign Affairs and the Arabian Peninsula Institute. Her work focuses on foreign policy, information warfare, security, and Middle Eastern affairs. She serves as Written Content Vice Chair of the American Bar Association’s Science and Technology Committee.

Scott Douglas Jacobsen and Irina Tsukerman discuss BRICS’ internal divisions, competing national interests, and limits on diplomatic cooperation. They examine artificial intelligence misuse, Anthropic’s findings, and weapons research, emphasizing distinctions between attempted misuse, operational capabilities, and speculation. The conversation addresses cybersecurity safeguards, technological proliferation, and the evidence needed for assessment.

Scott Douglas Jacobsen: Welcome back to Everywhere Insiders with Irina Tsukerman and Scott Jacobsen. Today’s discussion draws on Associated Press reporting about the BRICS summit in New Delhi.

BRICS leaders expressed concern about conflicts in the Middle East and Ukraine and opposed unilateral sanctions imposed without United Nations Security Council authorization. The original suggestion that Ukraine was not mentioned needs correcting. India’s Prime Minister Narendra Modi emphasized a greater role for developing countries in global governance.

The participants include countries with sharply different interests and relationships with Western powers. India, for example, seeks to maintain its relationships with Russia and China while strengthening partnerships with the United States and other Western countries. Those differences complicate the group’s efforts to act collectively.

What are your thoughts on the joint statement and the participants’ ability to cooperate diplomatically amid these conflicts?

Irina Tsukerman: In my view, BRICS has been a somewhat meandering bloc. Russia and China can use it to challenge Western influence, but describing the entire group as anti-Western overlooks the different priorities of its members.

The grouping originally brought together Brazil, Russia, India, and China, with South Africa joining later. Its members have promoted alternatives to financial arrangements dominated by Western countries, including payment systems that would reduce dependence on the U.S. dollar. That is an ambition, however, rather than evidence that they have displaced the dollar or removed themselves from the American financial system.

I would distinguish those financial ambitions from claims about a coordinated BRICS technology-transfer programme. AP’s summit coverage describes cooperation on artificial intelligence, energy, food security, trade in local currencies, and cross-border payments. Those initiatives do not establish that BRICS operates a collective mechanism for transferring weapons or quantum technologies.

The underlying difficulty, as I see it, is that the participants have diverse political systems, economic structures, and geopolitical agendas. Membership does not eliminate mistrust or competing national interests.

India and China illustrate that problem. Their leaders have sought to improve relations, but the countries continue to face border disputes and strategic competition. Cooperation within BRICS does not mean their objectives are identical.

Some members see the group as an opportunity to increase their economic and diplomatic flexibility. I would therefore describe BRICS as a forum for political and economic cooperation rather than a defence alliance. The Shanghai Cooperation Organization is a separate regional political, economic, and security grouping; calling it simply “the defence alliance” is misleading.

The question is whether BRICS can translate its declarations into concrete diplomatic achievements. In my assessment, statements alone are insufficient to establish its effectiveness as a mediator.

I think Xi would welcome a stronger diplomatic role for the group. Modi could also see value in that role despite India’s differences with China. Their interests could overlap on a particular initiative without producing broader strategic agreement.

For Russia, that prospect cuts both ways. A forum that gives Moscow diplomatic influence could also invite scrutiny of its war against Ukraine. My expectation is that Russia would support initiatives it considers useful while resisting language or mechanisms that expose its own conduct to greater criticism.

If anything, I think tensions within BRICS are likely to increase over time. In my assessment, Trump’s polarizing role has contributed to friction among Western allies as they work out how to respond to his policies. But disagreements within the West do not make BRICS a cohesive alliance.

Its members have different concerns and different ways of addressing them. India, in particular, balances its relationships with Russia and China against its partnerships with Western countries. AP’s reporting describes competing political, economic, and foreign-policy interests within the expanded group.

There are reasons to examine BRICS’ influence closely, but I think some of its messaging generates more attention than its practical achievements warrant. That messaging can have rhetorical value and help mobilize political supporters. Whether it produces coordinated industrial or diplomatic action is a separate question.

I would not say that its members have made no serious efforts to cooperate. My concern is whether they can implement shared commitments when their national priorities diverge. I expect similar difficulties in diplomacy because diplomatic engagement reflects individual countries’ objectives, political institutions, and approaches to negotiation.

Jacobsen: Turning to artificial intelligence, AP has reported on Anthropic’s findings concerning attempts to misuse Claude for weapons-related research, spyware, and propaganda. These are findings reported by the company, and attempted misuse should not be confused with the successful development of a functioning weapon.

This is not the first public reporting on the issue. In November 2025, AP reported that Anthropic said it had disrupted an AI-assisted cyber operation that its researchers linked to the Chinese government.

When you encounter a report like this, how should readers think about the interval between the underlying activity and its public disclosure? Could it be weeks, months, or years?

Tsukerman: I would be cautious about assigning a standard interval. These systems are developing quickly, and public reporting does not necessarily establish when an operation began, when it was detected, or how long investigators needed to assess it.

We also need to distinguish between a model providing information, assisting an operation, and controlling a weapon. Those are different capabilities, and evidence of one does not establish the others.

The dispute between Anthropic and the Pentagon illustrates that distinction. AP reported in February that Anthropic sought restrictions on mass surveillance of Americans and fully autonomous weapons. The Pentagon insisted on access for lawful uses without those contractual limitations. That disagreement does not establish that a particular autonomous weapon had already been deployed.

In March, AP reported that a Pentagon official connected the dispute to possible future applications in the proposed Golden Dome missile-defence programme. The discussion concerned potential capabilities and levels of autonomy, rather than proof of a first operational deployment during the Iran war.

Operational secrecy, investigations, and decisions about disclosure can affect when information becomes public. Without dates for a specific case, I cannot responsibly reduce that interval to a few weeks or months.

The Yemen example also requires precision. AP reported that Anthropic identified users in Houthi-controlled territory who attempted to use Claude to develop advanced missile technology. Their location did not establish their organizational affiliation, and the account described unsuccessful efforts, including a failed rocket test. It did not demonstrate that the Houthis had successfully developed an operational weapon using Claude.

Nor does that account establish that Iran, China, or Russia supplied access. Naming a state intermediary or adding another month to the presumed timeline would be speculation.

My assessment is that public disclosure can follow quickly when activity becomes visible, but visibility alone does not tell us when access began or whether the attempted application worked. Those questions require evidence from the specific operation.

I think the pace of technological development, implementation, and attempted theft will continue to increase. Security breaches and vulnerabilities, including those affecting AI systems, could make it easier for powerful capabilities to spread beyond their developers’ intended controls.

Anthropic CEO and co-founder Dario Amodei has called for slower AI development to allow safety measures to catch up. That broader appeal should be distinguished from the company’s dispute with the Pentagon over restrictions on mass surveillance of Americans and fully autonomous weapons.

My concern is that security safeguards and counterintelligence efforts could struggle to keep pace with attempts to misuse these technologies. However, we should distinguish documented misuse from claims that an AI system has escaped its controls or independently carried out lethal attacks.

AP reported that Anthropic identified attempts to use Claude for advanced missile development by users in Houthi-controlled Yemen. That account described unsuccessful efforts, including a failed rocket test. It does not establish that Claude enabled successful attacks on bases, soldiers, or ships.

We also need to be precise about self-improvement. Describing AI algorithms as “self-taught” does not establish that deployed systems independently add capabilities, eliminate vulnerabilities, or accelerate their own development without human intervention. The AP reporting discussed here does not substantiate those claims.

I expect increasingly capable AI tools to put pressure on the speed of both offensive activity and defensive responses. But that is an assessment, not evidence that espionage or public-disclosure timelines are shortening in a geometric progression. We still need to examine when each incident began, when it was detected, and when it became public.

Jacobsen: Thank you very much for the opportunity and your time, Irina.

—

Scott Douglas Jacobsen is a Writer and Editor for The Good Men Project with more than 2,000 publications on the platform. He is the Founder and Publisher of In-Sight Publishing (ISBN: 978-1-0692343; 978-1-0673505) and Editor-in-Chief of In-Sight: Interviews (ISSN: 2369-6885). He writes for International Policy Digest (ISSN: 2332–9416), The Humanist (Print: ISSN, 0018-7399; Online: ISSN, 2163-3576), Basic Income Earth Network (UK Registered Charity 1177066), Humanist Perspectives (ISSN: 1719-6337), A Further Inquiry (SubStack), Vocal, Medium, The New Enlightenment Project, The Washington Outsider, rabble.ca, and other media. His bibliography index can be found via the Jacobsen Bank at In-Sight Publishing comprised of more than 10,000 articles, interviews, and republications, in more than 200 outlets. He has served in national and international leadership roles within humanist and media organizations, held several academic, media, and research institution fellowships, and currently serves on numerous international boards. He is a member in good standing in numerous media organizations, including the Canadian Association of Journalists, PEN Canada (CRA: 88916 2541 RR0001), and Reporters Without Borders (SIREN: 343 684 221/SIRET: 343 684 221 00041/EIN: 20-0708028), and others.