Security
Malware in LiteLLM 1.107.0 steals cloud and AI keys
Teams that operated LiteLLM 1.107.0 must think that attackers have their secrets, and the last safe version is 1.106.2.
What matters in AI.
SubscribeTag
79 stories carry this tag.
Security
Teams that operated LiteLLM 1.107.0 must think that attackers have their secrets, and the last safe version is 1.106.2.
Agents
The plugin reports a regression only when a test on a live copy of the system shows it.
blog.trailofbits.comClaimed, not confirmed
Industry
rhino.ai says it modernized more than 500 undocumented Java procedures in nine months, against an initial plan of 18 months.
Security
Solcis says its Policy Engine gives each agent task an Allow, Deny or Escalate decision in less than 200 milliseconds.
pan african visionsClaimed, not confirmed
Security
An attacker can make an agent operate a tool with no model check. All 3 vendors made patches.
Agents
Coinbase tells that AWS and Cloudflare, which have about 40% of internet traffic, let publishers charge agents with x402.
finance.biggo.comClaimed, not confirmed
Agents
An analyst says that Context Engine is read only, but Neo4j says it gives agents a map of company data.
TechTargetClaimed, not confirmed
Industry
A new customer can sign in with a Google, GitHub, Apple or Amazon account, and most give no credit card.
Compute
For Qwen3-32B, the time for each output token also got better by a maximum of 30 percent, compared with the default pull mode.
Compute
In tests on Amazon GPUs, TTFT fell by 1.6 times for Qwen3-32B and 3.3 times for GLM-5-FP8.
Security
Perplexity reports that 108 attacks on the sandbox did not go to the host system.
Agents
In tests by AWS, Strands Decider 2B makes a decision in about 115 milliseconds on an NVIDIA RTX 3090.