GTG-1002: The New Face of Cyber Threats

At the center of the discussions was a recurring theme: GTG-1002, a cyber-espionage campaign detected in mid-September 2025 and attributed with high confidence by Anthropic to a Chinese state-sponsored group. According to the report published in November 2025, the operation targeted approximately 30 organizations across the technology, financial, chemical, and government sectors, using an unprecedented level of AI autonomy throughout the entire attack chain.

Beyond the striking nature of the case itself, GTG-1002 primarily acts as a wake-up call. It highlights a profound shift in cyber threats: attacks that are faster, more autonomous, and capable of exploiting the delays, blind spots, and inconsistencies of security systems still largely designed around human response times.

Here are the main takeaways from the summit.

1. The speed of attacks is now changing the very nature of risk

The first major observation: speed is no longer just an aggravating factor, it has become the core issue. According to Anthropic, the operators behind GTG-1002 reportedly used Claude Code to automate 80 to 90% of tactical operations, from reconnaissance and exploitation to lateral movement and data exfiltration. All at a pace described as “physically impossible” for a human to replicate.

“In practical terms, a complete attack takes less time than ordering a coffee,” summarizes Trevor Marshall, National Director at Delinea Canada.

This reality fundamentally changes the defensive equation. For years, organizations built their security mechanisms on the assumption that they would have at least some time to detect, analyze, qualify, and respond. But when attacks unfold at AI speed, that time window collapses dramatically. Risk is no longer measured solely by exposure surface or vulnerability level, but by the ability to observe, decide, and act fast enough.

That point surfaced repeatedly throughout the summit discussions: when attacks can automatically orchestrate reconnaissance, map environments, identify privileged identities, and chain actions together without meaningful pauses, sequential response models quickly reach their limits.

“When an attack becomes faster than Sonic or Flash, but our response still depends on a human, a coffee, and a change request, the outcome starts becoming statistically predictable,” joked Marcellin Nachin, Cybersecurity Leader.

An attack that moves faster than an organization’s decision-making cycle becomes, by definition, much harder to stop.

2. GTG-1002 shows that tools alone are no longer enough

The second major takeaway: stacking tools is not a resilience strategy. GTG-1002 clearly illustrates why it is dangerous to believe that a sophisticated technology stack alone is enough to protect the enterprise. Several speakers highlighted a familiar paradox: organizations often operate dozens of cybersecurity tools while still maintaining blind spots, operational friction, and gaps between theoretical controls and real-world conditions.

What GTG-1002 reveals is not only the power of an AI-orchestrated attack. It also exposes the fragility of security models that remain too fragmented.

The question is no longer simply “Do we have the right tools?” but rather “Is our security model still adapted to autonomous, distributed, ultra-fast attacks?” That means rethinking identity governance, real-time authorization, protection of non-human identities, reduction of unnecessary privileges, and the ability to eliminate the operational friction that slows teams down instead of strengthening them.

In other words, the response to AI-driven attacks will not be purely technological. It will also be structural, operational, and organizational.

The amplification caused by A2A (agent-to-agent)

One of the elements highlighted through GTG-1002 was the hijacking of legitimate internal agents exposed externally by organizations and subsequently exploited by attackers. The root cause? Poorly secured agents, often developed too quickly as proof-of-concepts whose success led to rushed production deployments. The lack of governance and proper identity and access management for these agents is often the real underlying issue. This is also a theme explored in our white paper.

3. Cybersecurity must be treated as a transformation issue, not just a technical function

GTG-1002 is not simply a story about offensive security or technical sophistication. It reveals a paradigm shift that directly affects governance, risk management, and transformation priorities.

At several points during the summit, the discussions raised simple but critical questions: Does the emergence of this new type of attack fundamentally disrupt cybersecurity strategy? Could our own internal agents become amplifiers for these attacks? How should organizations work with suppliers and partners to rapidly strengthen cyber capabilities? What priority should be given to secrets and privilege management?

These questions matter because they reposition cybersecurity where it now belongs: at the level of enterprise-wide strategic decision-making, with the goal of becoming more responsive during incidents and more resilient operationally. Achieving this also requires educating executives about this new threat landscape so cybersecurity can become a business enabler rather than simply a defensive function.

The GTG-1002 case clearly demonstrates that the boundaries between technology, operations, identity, governance, and strategy are becoming increasingly blurred. When agents can accelerate every stage of an intrusion, purely local or purely technical responses quickly hit a ceiling. Resilience then depends on the organization’s ability to align teams, clarify responsibilities, simplify control mechanisms, and rethink operating models.

4. The imperative is no longer to prepare “someday,” but to act now

If there is one shared conviction that emerged from the summit, it is this: the time to act is now. The debate is no longer whether agentic AI will transform the nature of cyber threats. That transformation is already underway. The real question is how quickly organizations are willing to evolve their own models in response.

Anthropic’s report presents GTG-1002 as a turning point: a campaign in which AI was used in unprecedented ways throughout the entire attack lifecycle, with human involvement limited to a few initialization and critical decision stages. Whether viewed as a warning sign, a precursor, or an accelerator, the conclusion remains the same: waiting for perfect maturity before acting is no longer an option.

That does not mean giving in to alarmism or launching disconnected initiatives. It means starting with the right priorities: identifying blind spots, reducing unnecessary privileges, strengthening truly critical controls, revisiting governance for both human and non-human identities, and above all, restoring speed in both decision-making and execution.

At its core, the most important lesson from GTG-1002 may be this: if attacks evolve at machine speed, inaction itself becomes a vulnerability.

GTG-1002 is not asking us to do better what we were already doing. It is asking us to operate differently. Believing we can complete decades-old unfinished security initiatives while simultaneously absorbing the rise of AI-driven attacks moving at machine speed is a comforting but dangerous illusion. It needs to be said clearly: we will not catch up by accelerating down the same road. The real question is no longer “How do we protect ourselves better?” but “What must we be capable of absorbing when an incident happens anyway?”

That shift is fundamental. It places resilience back at the center of strategy, not as an admission of failure, but as a realistic posture. An organization’s true capability is now measured by its ability to continue operating, making decisions, and recovering, not simply by avoiding disruption altogether. That means strengthening business continuity capabilities, shortening decision-making loops during crises, and embedding resilience far beyond IT teams alone. And thus, if attacks evolve at machine speed, inaction is a vulnerability, and resilience is a strategy.