Mahasiswa Bandung Temukan Celah Claude AI, Diganjar Rp66 Juta dari Anthropic

By Jurnalis Berita

🌍 Translate this article:

Thecuy.com reports that a student from Bandung has discovered a security flaw in Anthropic’s Claude AI system. For his discovery, he received a bug bounty reward of US$3,700, approximately Rp66 million.

The student is Muhamad Arga Reksapati, an informatics engineering student at the Bandung Institute of Technology (UTB). His finding was recognized by Anthropic after the verification process.

Arga’s Journey Into Cybersecurity Research

Arga’s path into cybersecurity was unconventional. He did not graduate from an IT‑focused school; instead he completed a vocational degree in mechanical engineering. Even in junior high, however, he was fascinated by how digital systems operate and how application security works.

He taught himself using online resources, covering fundamentals of coding, reading source code, building proof‑of‑concept (PoC) exploits, and joining international bug‑bounty programs. This interest drove him to pursue a degree in informatics engineering at UTB, where he continued to sharpen his cybersecurity skills.

Finding a Flaw in Claude Code Action

His expertise paid off when he uncovered a vulnerability in Claude Code Action, the mechanism that Anthropic uses to run AI assistants on GitHub Issues and Pull Requests. He submitted the discovery through Anthropic’s official bug‑bounty program hosted on HackerOne.

Following review by Anthropic’s security team, Arga’s report was confirmed as valid. He pinpointed a weakness in the data‑consistency guard that safeguards GitHub Issues and Pull Requests. Under specific circumstances, this protection can be bypassed via an image‑processing step, potentially allowing the AI system to read altered information after the initial processing stage.

AI Is Merely a Tool

Arga also used AI technology during his investigation, but it was only to speed up initial analysis and generate hypotheses. All verification, testing, and proof‑of‑concept steps were performed manually.

He cautions that indiscriminate use of AI in security research can lead to less accurate findings if not directly validated by the researcher.

Recognition from Anthropic

After the report was validated, Anthropic awarded a bug bounty of US$3,700 (roughly Rp66 million) as appreciation. The payment was sent via HackerOne, the official platform for Anthropic’s bug‑bounty program. Unlike several other bug‑bounty schemes, Anthropic does not issue a certificate of merit; instead, it provides a monetary reward that reflects the severity of the vulnerability.

Inspiration for Indonesia’s Young Talents

Arga’s achievement demonstrates that young Indonesian talent can compete on the global cybersecurity stage. Bug‑bounty programs have become standard practice at major tech firms worldwide, including Anthropic, Google, Microsoft, Meta, and OpenAI. Through these initiatives, independent security researchers can responsibly discover and report flaws for prompt remediation.

His success is hoped to motivate Indonesia’s youth to continue building expertise in technology, especially cybersecurity, which is now one of the most sought‑after skills in the digital era.

Baca juga Berita lainnya di News Page

Tinggalkan Balasan