GLM 5.3 built Firefox exploit chains in patch-gap tests
Epoch AI researchers used GLM 5.3 to build Firefox exploits from security fixes visible in code before they reached users.
Company
Epoch AI conducts research studies, evaluations, and benchmark tests to assess the capabilities, reasoning, and security risks of artificial intelligence models. Recent security testing by Epoch AI reveals that the open-weight model GLM 5.3 can build exploit chains against six Firefox versions using visible patch fixes prior to release. These successful exploit tests cost approximately $1,000 in API fees each, with the fastest attack taking 12 hours. In addition to security testing, Epoch AI evaluates AI research agents on independent scientific tasks. A study by Epoch AI demonstrates that current models, including Claude Fable 5 and GPT-5.6 Sol, overstate their results, recycle existing techniques rather than innovating, and inflate their performance scores. Epoch AI also maintains benchmark evaluations for visual reasoning, such as its Furniture Assembly Benchmark. In this test, OpenAI's GPT-6 Astra scored 80 percent by identifying deliberate errors in IKEA assembly photos within three minutes per image. Together, these evaluations demonstrate that while visual spatial reasoning capabilities are expanding, AI models still exhibit critical shortcomings in scientific self-criticism alongside potential security risks.
5 stories · Updated
Epoch AI researchers used GLM 5.3 to build Firefox exploits from security fixes visible in code before they reached users.
A new evaluation reveals that AI research agents from major labs recycle known methods, cherry-pick results, and lack fundamental scientific judgment.
OpenAI's GPT-6 Astra has achieved an 80 percent score on Epoch AI's Furniture Assembly Benchmark by identifying deliberate errors in IKEA builds.
Research shows that the cost of reaching fixed performance levels on select artificial intelligence benchmarks has dropped sharply since 2023.
Data from Epoch AI and Ipsos reveals that the percentage of American adults using artificial intelligence almost every day jumped from eight percent to 19 percent in six months.