research

Tencent Worked With International AI Safety Groups to Assess Risk of Its Hy4 Open-Weight Model

TL;DR

Tencent Holdings worked with an international group of AI safety organisations to assess the risk level of its latest open-weight model, Hy4. The South China Morning Post reports it is the first time a leading Chinese tech company has disclosed using independent third-party evaluators for a new model before release.

1 min read
0

Tencent Holdings has teamed up with an international group of AI safety organisations to assess the risk level of its latest open-weight model, Hy4, according to the South China Morning Post (SCMP).

The assessment comes amid growing concerns over AI-assisted cyberattacks and biological misuse, the two risk areas SCMP names in its report.

First disclosed third-party evaluation by a leading Chinese tech company

SCMP reports that the development marks the first time a leading Chinese tech company has disclosed that it worked with independent third-party evaluators in such an evaluation of a new model before releasing it.

The report quotes a Tencent statement that begins: "Our goal is to develop AI systems that are not only increasingly..." The rest of the quote was not available.

Details not yet available

Specifications for Hy4, including parameter count, context window, pricing and benchmark scores, are not yet available. The names of the participating safety organisations and the results of the assessment are also not yet available.

What this means

Tencent is publicly tying the pre-release risk review of an open-weight model to outside evaluators rather than only internal testing. The review targets cyberattack and biological misuse, the two concerns SCMP highlights. Because the model is open-weight, the risk assessment is a notable part of how Tencent presents Hy4.

Source: scmp.com ↗

Related Articles

research

inclusionAI releases SingProbe, an 8.13M-parameter streaming guardrail for Ling-3.1-flash-1019

inclusionAI published Ling-3.1-flash-1019-singprobe, an 8.13M-parameter probe that reads three hidden-state layers of Ling-3.1-flash-1019 and scores query intent, response unsafety and hallucination risk at each generated token. The model card reports a response safety F1 of 0.8707 and a hallucination AUC of 0.8041.

research

Anthropic mathematician claims Riemann Hypothesis proof with Claude; Lean community certifies it correct

Anthropic mathematician Levent Alpöge said on November 20 that he and Claude found a proof of the Riemann Hypothesis, published as a formal proof in Lean. On November 21, the Lean and Mathlib communities issued a joint statement certifying the proof as correct. No human-readable paper exists yet.

research

Nvidia team: huge context windows often overkill, as summaries and lookups matched them in many tests

An Nvidia research team argues that very large context windows are often overkill and expensive, according to The Decoder. In many tests, periodic summaries and targeted lookups worked about as well.

research

Bytedance-led team argues three training steps are enough for AI agents, releases competitive open model

A team from Bytedance and several US universities argues in a new paper that three training steps are enough to build capable AI agents. The authors say stacked, multi-phase pipelines make it hard to tell which step helps and make reproduction expensive. According to The Decoder, the work is backed by a competitive open model.

Comments

Loading...