ChatGPT 5.5 Pro launched this week. An AI research team found that LLMs corrupt documents when you delegate work to them. The EU wants to close VPN “loopholes.”
ChatGPT 5.5 Pro Quietly Goes Live
OpenAI released ChatGPT 5.5 Pro without the usual fanfare. Early user reports suggest significant improvements in mathematical reasoning and code generation. One Cambridge mathematician tested it on advanced problems and found it could handle complex proofs that stumped previous versions.
This matters because we’re seeing incremental but meaningful improvements in AI reasoning. Companies using AI for technical work — code reviews, data analysis, research — will likely see better results. But the quiet launch suggests OpenAI might be moving away from the hype cycle toward steady iteration.
For businesses running AI teams, this is exactly what you want: consistent capability improvements without the disruption of major model switches.
LLMs Mess Up Your Documents When You Delegate
A new study found that LLMs consistently introduce errors when you ask them to handle document tasks. The researchers tested common business scenarios: editing reports, summarizing meetings, formatting presentations. The AI models would subtly change facts, drop important details, or add information that wasn’t there.
This isn’t about hallucinations — it’s about delegation failure. When you hand off document work, you lose control over accuracy. The study showed error rates of 15-30% across different document types, even with careful prompting.
The takeaway: AI is great for drafts and suggestions, terrible for final documents you don’t personally review. At Kerios, this is why our autonomous teams always maintain audit trails — you need to see what changed and when.
EU Targets VPNs in Age Verification Push
The EU Parliamentary Research Service published a report calling VPNs “a loophole that needs closing” for age verification systems. They want platforms to detect and block VPN traffic from users trying to access age-restricted content.
This isn’t just about social media. Any business operating in the EU that serves different content based on location or age could face new compliance requirements. VPN detection technology exists but it’s expensive and creates friction for legitimate users.
For companies building AI systems that process user data, this adds another compliance layer. Geographic restrictions become harder to enforce, and user verification becomes more complex.
Try Kerios to see how autonomous AI teams handle compliance without the overhead.