[KongchangAI]
PaperSafe Personal RSI

Safe Personal RSI

A safety evaluation and repair process for recursive self-improvement at the prompt level of personal agents. Before an instruction update is adopted, both safety and task performance are checked, and fallback on failure and a repair feedback loop are supported.

Timeline (last 90 days)

Nov 19

Safe Personal RSI is an evaluation and repair pipeline component designed for personal agent prompt updates that checks both safety and task performance before adoption.

Unverified50%
Nov 19

The evaluation pipeline tests the proposed instructions and the current instructions separately, reruns failure cases to confirm stability, and adopts only versions that pass both safety and task performance.

Unverified50%
Nov 19

If no version passes evaluation, the current instructions remain in effect (falling back to the default policy of the old version).

Unverified50%
Nov 19

The executable benchmark covers 29 cases distributed across four scenario categories—email, calendar, health, and trading marketplace—and separately sets up probes to check authorization reuse, memory, and tool output injection.

Unverified50%
Nov 19

Across three models and two deliberately designed risky edits, the pipeline rejected 20 of 21 update attempts.

Unverified50%
Nov 19

One adopted GPT-5.6 Luna fix reduced attack violations from 10.7% to 0% and raised safe task completion from 89.7% to 100%.

Unverified50%
Nov 19

A single broad hand-written permission rule also achieves the same effect as the GPT-5.6 Luna fix (violation rate to zero, task completion rate 100%).

Unverified50%
Nov 19

The migration fix is effective but does not bring violations to zero; under strong attacks, runtime-level fallbacks are still needed, and there is an upper limit to the portability of prompt-based fix solutions.

Unverified50%
Nov 19

A standalone probe checked memory entry deletion and found a missing local OAuth scope check in the evaluated snapshot.

Unverified50%

All Facts (9)

Unverified

Safe Personal RSI is an evaluation and repair pipeline component designed for personal agent prompt updates that checks both safety and task performance before adoption.

50%
Unverified

The evaluation pipeline tests the proposed instructions and the current instructions separately, reruns failure cases to confirm stability, and adopts only versions that pass both safety and task performance.

50%
Unverified

If no version passes evaluation, the current instructions remain in effect (falling back to the default policy of the old version).

50%
Unverified

The executable benchmark covers 29 cases distributed across four scenario categories—email, calendar, health, and trading marketplace—and separately sets up probes to check authorization reuse, memory, and tool output injection.

50%
Unverified

Across three models and two deliberately designed risky edits, the pipeline rejected 20 of 21 update attempts.

50%
Unverified

One adopted GPT-5.6 Luna fix reduced attack violations from 10.7% to 0% and raised safe task completion from 89.7% to 100%.

50%
Unverified

A single broad hand-written permission rule also achieves the same effect as the GPT-5.6 Luna fix (violation rate to zero, task completion rate 100%).

50%
Unverified

The migration fix is effective but does not bring violations to zero; under strong attacks, runtime-level fallbacks are still needed, and there is an upper limit to the portability of prompt-based fix solutions.

50%
Unverified

A standalone probe checked memory entry deletion and found a missing local OAuth scope check in the evaluated snapshot.

50%

Source Articles